id
stringlengths 24
24
| idx
int64 0
402
| paragraph
stringlengths 106
17.2k
|
|---|---|---|
63d9c9be01ecc690f919ac6f | 53 | where N is the number of samples, k is the size of the neighborhood in the reduced space, U i is the set of points that are not neighbours in the original space but are now neighbors in the reduced space and r(i, j) is the rank of the sample j according to its distance from i in the original space . |
63d9c9be01ecc690f919ac6f | 54 | Steadiness is a measure of the loss of existing groups, and cohesiveness is a measure of the introduction of false groups. Briefly, the steps to calculate the steadiness and cohesiveness are are outlined below. We direct the interested reader to for further details. First, one computes the shared-nearest neighbor distance between points in the original space and in the projected space and constructs a dissimilarity matrix identifying compression and stretching of point pairs. Then average partial distortions are computed by randomly extracting clusters from one space and evaluating their dispersion in the opposite space. Once such partial distortions are known one can aggregate the results into steadiness and cohesiveness, |
63d9c9be01ecc690f919ac6f | 55 | where w i m compress|stretch i denote the iterative partial distortion measurements and their corresponding weights. Steadiness and Cohesiveness differ from trustworthiness and continuity in that they evaluate the authentic transformation of clusters For all models except the Transformer. all scores are >0.5 for all four metrics and the scores tend to hover around 0.75, with some exceptions. While the 128-dimensional Transformer performs best according to the steadiness score(∼ 0.8), the Transformer model in general performs worse on all other metrics than the other models, particularly cohesiveness, for which all three Transformer models perform worse than any other model, and in particular the 128-dimensional Transformer displays an exceedingly poor performance of just over 0.25. |
63d9c9be01ecc690f919ac6f | 56 | Overall we observe it is possible to construct linear projections of model latent spaces with comparatively low overall distortions; however, the model that has thus far performed particularly highly on other metrics (the 128-dimensional Transformer) has the greatest distortion, particularly in terms of cohesiveness, meaning it potentially introduces a significant number of false groupings. The Transformer's comparatively high distortion overall underscores one of the traditional trade-offs of machine learning: with greater power comes lessened interpretability. Now that we have an idea of how confident we may be in the representation of the PCs for the model latent space, we investigate the identification of "bridge variables" for further interpretation of the latent space. Bridge variables are known quantities of relevance in a scientific problem that show correlations with unknown variables, allowing heightened interpretability of many nonlinear problems . We consider a series of physicochemical properties of peptides that are measurable from sequence alone: Aliphatic Index, Boman Index, Isoelectric Point, Charge pH=3, Charge pH=7, Charge pH=11, Hydrophobicity, Instability Index, and Molecular Weight. We measure each of these properties for the test set sequences using the peptides python package . We employ the Silhouette Score in one dimension to measure correlation between individual PCs and verified-AMP/nonverified-AMP labeling, and we use the Pearson correlation coefficient (r) to measure correlation between individual PCs and the continuous physicochemical properties. |
63d9c9be01ecc690f919ac6f | 57 | The Aliphatic Index is defined as the relative volume occupied by the aliphatic side chains, Aliphatic Index = X(Ala) + aX(Val) + b(X(Ile) + X(Leu)), where X(Ala), X(Val), X(Ile), X(Leu) are mole percent of alanine, valine, isoleucine and leucine. The constants a and b are the relative volume of aliphatic side chains to that of alanine side chain where, a ≈ 2.9 and b ≈ 3.9. The Boman Index is equal to the sum of the solubility values for all residues in a sequence. It gives an estimate of the peptides' likelihood to bind to membranes or other receptors. Peptides with Boman Index > 2.48 are said to exhibit high binding potential. Isoelectric point Compiled by A. D. is the pH at which the net charge of a protein becomes zero. The charges at various pH levels Compiled by A. D. are determined according to the known isoelectric points of the amino acids. |
63d9c9be01ecc690f919ac6f | 58 | The hydrophobicity is a measure of the degree to which the peptide is hydrophobic, calculated by averaging the hydrophobicity values of each residue by using the scale proposed in . The instability index predicts whether a peptide will be stable in a test tube as presented in . The molecular weight is the average molecular weight of the peptide found by summing the individual masses of its amino acids and is directly correlated with sequence length. |
63d9c9be01ecc690f919ac6f | 59 | In Fig. , we identify the single PC with the highest correlation to each one of the potential bridge variables and illustrate the value of that maximal correlation. We note that although there is a comparatively low Silhouette score in one dimension indicating that the score in two dimensions is a more appropriate quantification of ordering in the system, we may use it to indicate which PC the models consider "most" relevant for verified-AMP ordering, and thus identify whether those PC's simultaneously correlate with physicochemical properties. For example, The 32-dimensional AAE, 64-dimensional WAE, RNN, Transformer and 128-dimensional AAE and RNN models employ one of the top 5 PC's most strongly for AMP ordering (Fig. ), and the same PC is also correlated with Charge pH=7 (Fig. ) and Isoelectric Point (Fig. ) for this model (Fig. ). |
63d9c9be01ecc690f919ac6f | 60 | For all models but the Transformer, the first principal component is highly correlated with molecular weight (Fig. ), which makes sense for the RNN, AAE, and WAE, as all are length-dependent models. That the RNN-Attention model also exhibits this behavior demonstrates the need to commit fully to a Transformer model to avoid a significant component of the model's variance being devoted to sequence length. The comparative performance of the AAE with the Transformer shows, however, that it is not necessary to remove the length dependence to enforce ordering capability in the model, as long as more than the first and second PC are considered. In a more general sense, any model with a high-variance component built in as part of the architecture will likely demonstrate this behavior. |
63d9c9be01ecc690f919ac6f | 61 | In these plots we find that the models will distribute the correlation across all of the first 5 PC's. Generally the transformer model is observed to make use of the first PC more often and this is expected to be because it does not feature a length based correlation in the first PC. |
63d9c9be01ecc690f919ac6f | 62 | Using these correlation coefficients it is possible to further characterise the ordering of the latent space and define directions along which interpolation should occur such that desirable characteristics will emerge from generated peptides. Each PC can be interpreted as a linear combination of the respective model's latent dimensions . From this interpretation it is possible to construct a direct relation between the latent dimensions of the model and the physicochemical properties investigated through the PCA mapping; we have made the PCA mappings themselves available on the project Github repository, . |
63d9c9be01ecc690f919ac6f | 63 | Our pipeline consists of three of the best performing models. We suggest the simultaneous exploration of the Transformer-128 (high reconstruction accuracy, good clustering separation in two dimensions, linear generated sequence partitioning and low interpretability), AAE-128 (moderate reconstruction accuracy, good clustering separation in two dimensions, non-distorted PCA space linear generated sequence partitioning, moderate linearity, medium interpretabilitybridge variables identified for AMP PCs), and WAE-128 models (moderate reconstruction accuracy, moderate clustering separation in two dimensions, moderate distortion in PCA space, linear generated sequence partitioning, moderate linearity, high interpretability). |
63d9c9be01ecc690f919ac6f | 64 | We then proceed to choose variables of interest to hold constant in the latent space and perturb the others. This will allow us to find new peptides with the properties of interest maintained. We choose to maintain the verified-AMP-correlated PC and the hydrophobicity-correlated PC, to produce the sequences likelier to share experimental AMP-likeness and hydrophobicity-an important AMP property-with GL13K. We note that this procedure could be done with any identified property near any input point of interest. We use the previously mentioned correlation coefficients (Sect. 3.4), to find the verified-AMP correlating PC and the top hydrophobicity correlating PC. Having kept track of the PCA vector for our AMP of interest, we perform sampling by keeping our two PC's of interest, AMP and hydrophobicity correlated, fixed and adding gaussian noise to all other PC's. The added noise variance must be tuned individually for each model as the latent space organization is different for each model. The variance tuning is performed by identifying the mean and standard deviation of all training dataset points in PCA space and then sampling Gaussian noise centered at the mean with 1/5 the standard deviation. This noise vector is then summed with the PCA vector of our AMP of interest, thus shifting the sampling location to be near the AMP of interest. Once we have the noisy nearby PCA samples we perform an inverse PCA transformation returning to our latent space vector representation, demonstrating one particular valuable property of the PCA approach. We then pass the noisy latent samples to the decoder which will generate the desired candidate peptide sequences. |
63d9c9be01ecc690f919ac6f | 65 | We demonstrate the use of the pipeline on GL13K. For each model, we generate five different sequences and display their properties (Figs. ). In the ideal case we should observe little change in Hydrophobicity of the samples while the other physicochemical properties should vary and in general this is what we observe. We observe that certain properties of random sequences generated in the neighborhoods tend to remain more constant irrespective of model (Aliphatic Index, Charge at pH=3, Molecular Weight, Isoelectric Point) and certain properties tend to be more variable (Boman index, Charge at pH=11, Instability Index), while certain properties are more dependent on the model. Hydrophobicity varies the most in the AAE model, and the least in the WAE model. One benefit of employing multiple models is the ability to sample different local neighborhoods of GL13K, with potentially different properties; another benefit is including both more interpretable and more high-performing models to generate samples. |
63d9c9be01ecc690f919ac6f | 66 | The results from the pipeline for the AAE-128, Transformer-128 and WAE-128 presented in the supplementary figures (Fig. . S8) show that by locking PC's of interest and sampling nearby points along other PC's we can generate novel peptides that have similar properties to the original GL13K. While some generated samples feature drastic changes in certain properties such as the fourth peptide from the AAE-128 with a charge of 1 at pH 11 or both generated peptides with -1 charges at pH 11 for the Transformer-128 and WAE-128 models, most properties fall near the original GL13K sequence properties. |
63d9c9be01ecc690f919ac6f | 67 | We have trained five deep learning models with VAE-like latent spaces on a library of sequences of short proteins and assessed the characteristics of the resultant models, including a careful examination of the properties of the latent spaces associated with them. We show that the models have the capability to create smooth, continuous latent spaces that can be sampled for de novo AMP generation and optimization. The inclusion of a simultaneously-trained property prediction network disentangles verified-AMP and non-verified-AMP sequences; however, to our surprise we find that, unlike in previous studies, this is not always apparent in the first two principal components derived by PCA. Different models associate different information with highly-varying PCs, and different models vary drastically in the ways in which they encode variance, and, hence, information. It is important to note that this presents a challenge for the incorporation of domain knowledge, since we see that the model does not necessarily place greater emphasis on user-provided information. Furthermore, it sounds a note of caution in the interpretation of latent space orderings, since observed orderings may occupy only a small fraction of the informational content in the model latent space. We have also addressed the question of how meaningful the use of PCA is as a tool for indicating properties of VAE-like latent spaces and argued that the less distortion imposed by PCA upon the different neighborhoods and interactions between points in the manifold, the more clearly interpretable the latent space is. The analysis that we present here may be applied to other short peptides of interest, such as anti-cancer peptides. Based on our results, we would suggest retraining our Transformer, AAE, and WAE models in conjunction with a new cancer/anti-cancer property predictor that relates to the property of interest. One could then employ a similar analysis and pipeline to assess the quality of the resultant latent spaces and generate sequences of interest. |
63d9c9be01ecc690f919ac6f | 68 | We have investigated the use of the principal components with the highest clustering of verified AMP properties for de novo AMP generation and showed that our models generate highly diverse and unique sequences, with comparatively low sequence similarity in local neighborhoods. Despite the low similarity and the use of a predictor trained on experimentally-verified AMP properties rather than direct knowledge of AMP-like-ness, all models but the 32 and 128-dimensional RNN-Attention and 64-dimensional AAE are capable of successfully partitioning a single coordinate in the latent space into regions that generate AMP-like sequences with high probability and regions that generate non-AMP-like sequences with high probability. We observe that the capacity of the model to reconstruct input sequences is not clearly linked to its ability to partition the space, and we add our voices to a number of cautions against the over-use of reconstruction accuracy as a metric for generative models. |
63d9c9be01ecc690f919ac6f | 69 | We have evaluated the extent to which models order their latent spaces according to standard peptide physicochemical properties that they are not trained on and identify the principal components most strongly correlated to given properties. We find that the models will order any of the first five principal components investigated according to physicochemical properties but when a model needs to assign a larger proportion of the variance to learning peptide length the first component is usually correlated with length / molecular weight. Indeed, only the 128-dimensional Transformer eschews length as a consideration almost entirely (no length dependence observed in the reconstruction accuracy, small correlation between any top five PC and molecular weight). The Transformer in general clearly has the capacity to function independent of the length but demonstrates a more rapid drop in performance as the latent space dimensionality is decreased than the RNN, WAE, or AAE. We speculate that this is due to the nature of the task being that discriminating between lengths can actually discourage models from overfitting; that is, from simply "memorizing" the answers, and may encourage a more meaningful lower-dimensional representation, although we also note that the 128-dimensional Transformer shows by far the most heightened local fragment-based similarity in its verified-AMP-relevant PC. This could suggest a model relying on fundamentally different information from the others. |
63d9c9be01ecc690f919ac6f | 70 | In terms of other relevant physicochemical variables, we observe moderate correlations (0.35 ≲ rPearson ≲ 0.65) between at least one PC and isoelectric point / charge at pH 7 in the AAE-64, RNN-32, RNN-64, Transformer-64 and all WAE models and moderate correlation between at least one PC and hydrophobicity in all but the RNN-Attention-32. As these are traditional hallmarks employed for AMP design, this is desirable behavior, and in particular aids in the interpretability of the models through a linear mapping of the latent space variables to straightforward "bridge variables" for most models. It also shows that the models are capable of identifying relevant properties from sequences alone, despite being trained only with a binary property predictor. |
63d9c9be01ecc690f919ac6f | 71 | We may further employ the bridge variables in conjunction with the probability of AMP generation in a single PC (Fig. ) to aid in the interpretability of the models. In our case models demonstrating a monotonic linear increase in probability, especially those reaching a prediction probability of > 0.6 (128-dimensional AAE, 64-dimensional RNN, 64-dimensional RNN-Attention, 64 and 128-dimensional Transformer and 32 and 64-dimensional WAE models) are arguably the easiest to interpret, since we now essentially have a single linear mapping from the latent space to AMP probability, which for the RNNs and WAEs are comparatively non-distorted from the original space (cf. Sec. 3.4). |
63d9c9be01ecc690f919ac6f | 72 | We demonstrate a trade-off between model complexity and model interpretability under this paradigm and suggest that for optimized design of AMPs in a continuous latent space, it may be appropriate to perform the optimization in multiple different latent spaces, using a similar philosophy to that of ensemble voting. We do a short case study to show one way this might be implemented, and indeed, in future work we plan to use this as a starting point for an active learning procedure to traverse these spaces and perform multi-scale molecular dynamics simulations upon relevant points. |
63d9c9be01ecc690f919ac6f | 73 | In the future, we plan to investigate a phenomenon termed selective latent memories due to Kullback-Leibler Divergence constraining. This effect is observed during training and causes a drop in the entropy of certain latent dimensions when the KLD is minimized. In addition, we will generalize from a binary AMP/non-AMP classifier to a multi-class predictor capable of grouping sequences by expected mechanism of action. Finally, as a prelude to the previously mentioned active learning traversal of the space, we plan to investigate the incorporation of structural data into the models, perhaps leveraging the recent success of AlphaFold2 and similar structural prediction algorithms. |
63d9c9be01ecc690f919ac6f | 74 | The models developed in this research used deep learning to discover embeddings for sequences of amino acids but future work should investigate other peptide representations such as protein structure distance graphs which can embed structural information and SMILES strings used in the world of small drug design. SMILES strings encode chemical information into a sequence of characters thus allowing the models to learn chemical distinctions between the amino acids. |
63d9c9be01ecc690f919ac6f | 75 | Overall, we have performed a thorough qualitative and quantitative analysis of the latent spaces of five different generative models for AMP design, identifying strengths and weaknesses of each, as well as developing a suite of analysis methods for latent space design and sampling in the context of generative deep learning of AMP sequences. We provide a much-needed set of benchmarking protocols in this nascent area of research. |
63da867133b6976c48031bbe | 0 | Drug discovery requires a significant amount of time and money, as it takes an average of 10 to 15 years and $2.6 billion to produce a new drug. Drug discovery research is initiated because there is a disease or medical condition for which no suitable drug is available. The hypothesis that inhibiting or activating a certain protein or pathway will have a therapeutic effect on a certain disease is formulated and verified. In general, hit-to-lead is an intensive SAR investigation around the structure of each core compound with measurements to establish the activity and selectivity of each compound while using methodologies such as molecular modeling, X-ray crystallography, and NMR to establish the magnitude of activity and selectivity of each compound. The optimization of lead compounds yields candidate compounds that undergo nonclinical and clinical trials, and are finally marketed after obtaining government approval. Lead optimization involves modifying known molecular structures to improve various parameters such as biological effects, physicochemical properties, toxicity, and stability. Typically, 200,000 to 1 million compounds are screened first. Then, more than 100 compounds are screened in hit-to-lead and lead optimization to narrow the molecules down to one or two candidates. Subsequently, it was shown that approximately 1 in 10 (10.4%, n=5,820) of the lead and all indications that entered Phase 1 were approved by the FDA. Thus, the project often fails and has a low success rate. If even a portion of these processes could be assisted by in silico methods, it would save a tremendous amount of time and expense. With recent developments of computers and algorithms, the application of computer science technology to drug discovery has been studied, and the efficiency and quality of the drug discovery processes have been improved. Molecular generative models are computer-based methods related to hit compound discovery and lead optimization. Molecular generative models have the advantage of being able to efficiently explore a huge chemical space and generate novel compounds with desirable properties using machine learning. They also have the advantage of avoiding explicitly dealing with complex chemical knowledge by using large compound datasets to train machine learning models. Several methods have been proposed to optimize molecules according to evaluation functions. The variational autoencoder can be used to generate molecules by modeling in the latent space. Then, optimization can be performed with the gradient method, leveraging the fact that the latent variables are continuous with the model that predicts the evaluation value from the latent variables. Such approximate models of the evaluation function are generally used in molecular optimization. String representation of molecules (SMILES) can be generated by long short-term memories (LSTMs) of recurrent neural networks . In this case, optimization is performed by retraining the generative model using the dataset in which molecules without the target property are removed from the generated molecules by the approximate model of the evaluation function. |
63da867133b6976c48031bbe | 1 | Optimization using generative adversarial networks (GANs) is performed in the same manner. Some optimization methods are also based on reinforcement learning. One popular approach is to represent molecular generation as a Markov decision process in which the state is a molecule and the action is the addition of atoms (or fragments) to approximate the policy function with a machine learning model. This approach has the advantage of directly optimizing the evaluation function, compared to using an approximate model. |
63da867133b6976c48031bbe | 2 | In addition to the policy gradient method, there are other optimization methods such as Q-learning and Monte Calro tree search (MCTS). Most molecular generative models focus on generating molecules from scratch and are not suitable for starting generation from a given arbitrarily molecule. While this is suitable for generating compound libraries used to search for hit compounds, it is not suitable for cases such as lead optimization, where a candidate compound has already been narrowed down and optimization is performed on that compound. Several methods focusing on starting generation from arbitrary molecules have been proposed. Mol-Cycle-GAN is one of the methods that starts generation from a given molecule. The method is trained using sets of molecules before and after optimization based on the Cycle-GAN generation scheme. Optimization is performed via latent variables, but according to experimental results, it is inferior in terms of performance compared with methods using reinforcement learning. MERMAID is the most relevant method for this study, and it starts generation from an arbitrary molecule by editing SMILES with MCTS and LSTM. However, the generated molecules often deviate significantly from the initial molecules as the optimization progresses. In this study, we developed a molecular optimization method based on molecular graphs, starting from an arbitrary molecule to be explored by MCTS. The use of molecular graphs allows for high similarity with the starting compound, while MCTS allows for efficient generation without prior learning of a specific evaluation function. In addition, a graph neural network model trained on the compound dataset is used to enhance the efficiency of the search. The search is conducted per fragment, but the fragments to be added are generated atom by atom. This allows more appropriate fragments to be added to the current molecule, while avoiding the lack of diversity caused by using a fixed fragment vocabulary. |
63da867133b6976c48031bbe | 3 | Table shows that the QED value was more than 0.92 for the top few cases, which is approximately sufficient for optimization. Furthermore, MERMAID had a lower evaluation function value of 0.3 to 1.4. This is also visualized in Fig. . The difference between the two methods is the representation of molecules: MERMAID uses SMILES, and the proposed method uses molecular graphs. Considering the ZINC dataset as an example of "data size," the average number of nodes in a molecular graph was approximately 24, while the average number of tokens in SMILES was approximately 40. Since both methods search only one node/token per step, this method was considered to search more when compared with the same number of steps. Additionally, while SMILES generated invalid SMILES, the molecular graph was always valid, which is another reason this method was able to search more. Note that SMILES with RNNs generally runs faster than molecular graphs with graph convolutional networks (GCN) when compared in a single step. |
63da867133b6976c48031bbe | 4 | In terms of validity, this method employing molecular graphs always produced valid molecules (i.e., validity = 1), while MERMAID using SMILES produced strings that do not satisfy the SMILES grammar (i.e., validity = 0.7). Novelty and uniqueness were all 1 or nearly 1, confirming that there were no problems when viewed from these perspectives. The similarity shows higher values for this method clearly (Fig. ). Given that the similarity for all pairs of molecules in the ZINC dataset of approximately 250,000 molecules was 0.144, the molecules generated by this method can be considered to be similar to the starting molecules. The SA score was clearly better for this method. This may be due to the fact that the molecular graphs used in this method allow only the addition and deletion of single bonds, making it difficult to create complex structures. |
63da867133b6976c48031bbe | 5 | Examples of molecules generated by this method are shown in Fig. . It can be seen that the addition and deletion of a few fragments from the starting molecule resulted in a molecule with improved QED values, while maintaining a high degree of similarity. One of the reasons for the higher similarity obtained with this method compared with MERMAID using SMILES is that the addition and deletion of fragments is limited to single bonds, thereby avoiding the direct editing of the ring structure as in MERMAID. However, this restriction of actions has a disadvantage in terms of diversity because complex rings cannot be generated (in the fragment-wise search). |
63da867133b6976c48031bbe | 6 | Table shows that the improved value of PLogP was slightly less than GraphAF but was still sufficiently better than that of the other three methods. The success rate was approximately 1, indicating that PLogP can be optimized regardless of starting molecules. Note that the comparison of the methods here is not completely fair because they have different generation schemes. For example, GraphAF and GCPN generate molecules after fully optimizing the policy model through reinforcement learning, whereas MERMAID and the proposed method exhibit a disadvantage in that they do not train the network for a specific evaluation function. |
63da867133b6976c48031bbe | 7 | We compared and evaluated the initial molecule and generated molecule in terms of the similarity given by the Tanimoto coefficient based on the ECFP4 fingerprints, except for the value of the evaluation function. However, in practice, lead optimization is expected to maintain not the similarity but also the potency and selectivity to the target. Although similarity may be correlated to these properties, it is more appropriate to evaluate in terms of the potency and selectivity. Considering the comparison of the initial molecule and generated molecule by such a metric for drug efficacy, this method is not expected to retain a high percentage of the original drug efficacy. The search space of this method is narrowed down to a molecule-like graph by the GCN model, and biased by MCTS toward regions with higher values of the evaluation function. Thus, if specific properties such as drug effects are to be considered, they must be explicitly handled in the evaluation function. Masking the evaluation function to preserve important substructures with respect to the property of interest is also effective. |
63da867133b6976c48031bbe | 8 | Most molecular generative models do not consider starting from a given molecule, and are not appropriate for situations such as lead optimization. Additionally, existing methods that start generation from a given molecule have problems such as low similarity to the starting molecule. Therefore, in this study, we developed a molecular optimization method that starts generation from arbitrary molecules based on molecular graphs. Optimization is performed in a fragment-wise tree search according to a given evaluation function. Fragments are generated individually by MCTS in an atom-by-atom manner. The GCN model trained on the fragment dataset is used to improve the efficiency of fragment generation. In an experiment of optimizing QED as an example of the evaluation function, the generated compound not only improved the value of the evaluation function sufficiently but also had a high similarity to the starting compound. In addition, it was confirmed that this method searches near the starting compound compared to existing methods. Thus, this method is considered to be suitable for processes such as such as lead optimization, where compound candidates have already been obtained. Furthermore, this method can be used to mask important structures identified beforehand, such that they remain unchanged. |
63da867133b6976c48031bbe | 9 | The molecular generation method developed in this study is a molecular graph-based optimization method for a given arbitrary molecule based on an evaluation function. The method performs a tree search with the molecule to be optimized as the root, by adding and deleting fragments based on the upper confidence bound (UCB) score. Fragments to be added are generated atom by atom with MCTS, and the efficiency is enhanced using a GCN model. Generating fragments atom by atom allows the molecule to be further optimized locally (only partial conformational changes). |
63da867133b6976c48031bbe | 10 | MCTS is a model-based reinforcement learning algorithm, and has been used as an effective policy improvement operator in deep reinforcement learning methods such as AlphaGo. MCTS searches the optimal sequence of state and action sequences by sequentially constructing a search tree with the initial state as the root. Each node has a state value and a number of visits, each initialized with 0. The following four steps are repeated as one cycle until a given convergence condition is satisfied. |
63da867133b6976c48031bbe | 11 | where x is the average reward for the self node n, and n p is the number of visits to the self node and the parent node. Default Policy is guaranteed to converge to the optimal solution with a sufficient number of steps even with random selection, but this is not feasible in practice. Machine learning models have been often used in recent years to search efficiently. |
63da867133b6976c48031bbe | 12 | GCN is a neural network comprising layers that perform convolution operations defined over a graph data. Convolution cannot be simply applied to graphs, unlike images or series in which the relationship between neighboring elements is not fixed. There are two types of convolutions on graphs: one dealing with signals over graphs and the other based on the spatial structure of graphs. Here, we describe graph convolution defined on the space used in this study. The output of the l layer of node i depends only on its neighbors, and is defined as follows: 42 |
63da867133b6976c48031bbe | 13 | where h l/l+1 is the output of thel/l + 1 layer, b l is the bias of the l layer, W l is the weight parameter, N (i) is the neighbor of the node i, c ji is the product of the order roots of the nodes j and i, and σ is the activation function. |
63da867133b6976c48031bbe | 14 | The fragment-wise search (Fig. ) is the core process of the method. Optimization is performed by editing the molecular graph fragment by fragment, given an arbitrary molecule and an evaluation function as input. The state of each node in the search tree corresponds to a molecule with a state value and a number of visits. The action in tree search is to remove and add fragments. The tree search with the starting molecule as the root searches with the following process as one cycle. The molecule corresponding to the newly added node is treated as the optimized molecule. |
63da867133b6976c48031bbe | 15 | The number of child nodes for removal is finite and small, e.g., an average of approximately 9 in the ZINC dataset, 43 but the number of child nodes for addition is significantly large due to the number of molecules generated by the fragment generation module. Therefore, the selection of some molecules based on some criteria is necessary. In this study, random selection and a criterion for a high value of the evaluation function, such as ϵ-greedy, 44 were selected. |
63da867133b6976c48031bbe | 16 | The evaluation value for the molecules of the newly added child node is used as the reward. The reason for this is that unlike board games where the terminal state can be easily determined, it is difficult to perform a rollout in molecular generation due to the ambiguity of the terminal state. For example, the benzene ring is non-terminal when naphthalene is generated, but it is terminal when benzene ring is generated. Existing methods use machine learning models to determine if the state is terminal, or to fix the number of steps. However, when the search starts from an arbitrary molecule and proceeds in the direction of increasing and decreasing atoms, as in this method, determining whether the search is terminated is difficult, because the initial molecule is already completed. Another reason is that deep nodes are likely to deviate significantly from the initial molecule, which is not in accordance with the purpose of the study. |
63da867133b6976c48031bbe | 17 | For all nodes in the path from the root node to the selected node, the state value and number of visits are updated using the maximum value of the rewards calculated in the Simulation step. The following transformations are applied to keep the reward value in the range of -1 to 1. |
63da867133b6976c48031bbe | 18 | Atom-wise search corresponds to the generation of newly fragmented molecules in the Expansion step of MCTS in the fragment-wise search (Fig. ). One path corresponds to one fragment by assigning one atom to one node. In one step, a new atom and the bonds associated with that atom are predicted to be added to the fragment in the intermediate state. Atoms that cannot be further bonded are excluded, considering the valence rule in the Expansion step. Atom-wise MCTS is executed as one cycle of the following process. |
63da867133b6976c48031bbe | 19 | Rollout is performed using the same GCN model used in the Expansion step to evaluate the added child nodes. Unlike the fragment-wise search, the fragments are generated from scratch; therefore, the terminal state can be defined in the same way as for existing methods. A state is treated as a terminal if the GCN model predicts an empty atom or when no bonds are predicted. The generated fragments are added to the molecule selected in the Selection step of MCTS in the fragment-wise search to obtain a new molecule, and the value of the evaluation function for that molecule is used as the reward. |
63da867133b6976c48031bbe | 20 | The atom prediction module takes the hidden state of the graph as input, and outputs the type of atom as a probability by passing it through the fully connected layer (F C a (•)). The dimension of the output is (atom type) +1, which is the sum of the number of atom types and the label indicating the termination. |
63da867133b6976c48031bbe | 21 | The bond prediction module takes the predicted atom and hidden states of the nodes as input, and predicts the bonds through the RNN layer. The initial state vector s of the RNN is the concatenation of the vector obtained by transforming the predicted atoms with the Embedding layer (Emb(•)) and the hidden representation of the graph h g . The input of the RNN is the hidden state vector of nodes h n arranged as a series of data according to the BFS order of nodes in the input graph. The output is the type of bonds as a probability, whose dimension is (number of bond types)+1 including a label indicating that there is no bond. s = concat(h g , Emb(y a )) |
63da867133b6976c48031bbe | 22 | The atom prediction module comprises two fully connected layers. The hidden layer has 64 dimensions and uses ReLu functions as activation functions. A Softmax function is applied to this output y a to obtain the probability of each atom. The output is ten-dimensional, comprising an empty atom meaning termination and nine types of atoms. |
63da867133b6976c48031bbe | 23 | The bond prediction module comprises a two-layer GRU 45 followed by a two-layer fully connected layer. The initial state vector of the GRU is a concatenation of the graph hidden state vector h g and the 64-dimensional embedded representation of the predicted atoms Emb(y a ), and is converted to 256 dimensions in a single fully connected layer. The output of the GRU is transformed to y b by the fully connected layer with the hidden layer having 64 dimensions and ReLu function as the activation function. The dimension of outputs is four, including the label indicating no bond. The probability of a bond is obtained by applying a Softmax function. |
63da867133b6976c48031bbe | 24 | The ZINC database used for training is a database for virtual screening, and contains over 750 million molecules. Approximately 250,000 molecules, which are commonly used in molecular generative models, were used for training. This dataset is the same as that used in ChemTS, and is publicly available at . 22,234 fragments obtained by applying the procedure described in the Methods section to each molecule were used for training. An example of the fragments used for training is shown in Fig. . |
63da867133b6976c48031bbe | 25 | The labels are atom type and bond type, and each loss was calculated as a cross entropy loss function where the sum of losses is the overall loss. The ratio of train/test data was 4:1, and parameters were updated with the Adam optimizer. The model was trained in a teacherforcing 46 manner, in which label data were used as inputs when there is a dependency between the inputs. The input for the bond prediction was not the output of the atom prediction module but the label of the atom, because the prediction of the bond depends on the prediction of the atom. The other training settings were as follows. |
63da867133b6976c48031bbe | 26 | In this experiment, optimization based on a single evaluation function was performed. The evaluation function was QED. The reason QED was used here is not that it is sufficient to increase QED in lead optimization but that it is easy to calculate, and QED is commonly used in the evaluation of molecular generative models. |
63da867133b6976c48031bbe | 27 | One hundred molecules with QED in the range of 0.6 to 0.7 were randomly selected from the ZINC dataset and used as starting molecules. The properties of the starting molecules used in this experiment are shown in Table2. For each molecule, a 20-step fragment-wise search was performed. Then, the atom-wise search was performed 100 steps for each molecule. As a result, the optimization was performed in 2,000 steps. |
65153bcaa69febde9ed55d25 | 0 | Diamond is the hardest known material with a Vickers hardness (H V ) at the level of 100 GPa and density ρ = 3.635 g/cm 3 . Its structure (space group Fm-3m) is formed by corner-sharing tetrahedra of sp 3 -hybridized carbon atoms with C-C-C = 109.47° and is characterized by the highest atomic density (i.e., the number of atoms per unit cell volume) and the highest density per valence electron . A rare hexagonal form of diamond (lonsdaleite) (space group P6 3 /mmc) with virtually the same density has been claimed to be stronger and stiffer than diamond . |
65153bcaa69febde9ed55d25 | 1 | The network topologies of diamond and lonsdaleite are dia and lon, respectively, and many theoretically predicted carbon allotropes have been identified with these topologies (see and references therein). The topology determination for the new phases is now made easy with the TopCryst program . Information on all carbon allotropes extracted from the literature is indexed in the "SACADA" database , which currently contains 703 allotropes. |
65153bcaa69febde9ed55d25 | 2 | Diamond (both cubic and hexagonal) is still considered to have superior atomic density, elastic moduli and hardness . Recently, however, several superdense (ρ > 3.635 g/cm 3 ) carbon allotropes have been predicted from the first-principles studies , and hexagonal (space group P 6 222) C 3 with assigned quartz topology (SACADA qtz #11) has even been claimed to have a hardness of 113 GPa , i.e., 15% harder than diamond. Also, very recently superdense (ρ = 3.666 g/cm 3 ) ultrahard (H V ≈ 102 GPa) hexagonal C 6 (space group P6 5 22) allotrope with qtz topology was proposed by us . |
65153bcaa69febde9ed55d25 | 3 | The structures of the binary compounds have all been subjected to geometry relaxations of the atomic positions and lattice constants down to the respective ground states characterized by minimum energies. The protocol consists of iterative calculations performed using the DFT-based plane-wave Vienna Ab initio Simulation Package (VASP) . For the atomic potentials, the projector augmented wave (PAW) method was used . The exchange and correlation effects were treated within a Generalized Gradient Approximation (GGA) scheme . The relaxation of the atoms to the ground state geometry was done by applying a conjugate-gradient algorithm . A tetrahedron method with corrections according to the Methfessel-Paxton scheme was used for geometry optimization and energy calculations, respectively. A special k-point sampling was applied to approximate the reciprocal space Brillouin zone (BZ) integrals. For better reliability, the optimization of the structural parameters was carried out along with successive self-consistent cycles with increasing k-mesh until the forces on atoms were less than 0.02 eV/Å and the stress components were below 0.003 eV/Å 3 . The plane waves energy cutoff was 400 eV. |
65153bcaa69febde9ed55d25 | 4 | The mechanical properties were derived from the elastic constants calculations . The phonon dispersion band structures were calculated to verify the dynamic stability of the new phases. The phonon modes were computed considering the harmonic approximation by finite displacements of the atoms around their equilibrium positions to obtain the forces from the summation over the different configurations. The phonon dispersion curves along the direction of the Brillouin zone were then obtained using the "Phonopy" interface code . The crystal information files (CIF), the structure sketches including the tetrahedral representations as well as the illustrations of the charge density plots were generated using the VESTA graphics software . The electronic band structures and densities of states were obtained with the full-potential augmented spherical wave ASW method based on DFT using the same GGA scheme as above . |
65153bcaa69febde9ed55d25 | 5 | The structure of qtz C 6 is shown in Fig. . A differentiation of carbon into two different sites was highlighted by resolving the initial C 6 structure (space group P6 5 22, No. 179) with unique (6a) sites into carbon atoms with (3c) and (3d) Wyckoff positions within space group P6 4 22, No. 181 as shown in Table . Such an elementary modification allowed to consider the binary compounds B 3 N 3 and Si 3 C 3 . The lattice parameters of the ground state structures are given in columns 3 to 5 of Table . The corresponding crystal structures are shown in Figs. with ball-and-stick and tetrahedral representations, the latter being characterized by corner sharing irregular tetrahedra (vide infra). The atoms are described with general Wyckoff positions, i.e., at (3c) ½, 0, 0 and at (3d) ½, 0, ½. Differences are observed in the volumes (total and atom-averaged) and in the interatomic distances, which increase along the series due to the increase in the respective atomic radii. If we look at the angles related to the constituent tetrahedra, they differ significantly from the regular tetrahedral one (109.47°), thus indicating the specificity of the qtz topology. |
65153bcaa69febde9ed55d25 | 6 | The analysis of the mechanical behavior was carried out using the elastic properties by performing finite distortions of the lattice. The phase is then described by the bulk (B) and the shear (G) moduli obtained by averaging of the elastic constants. Here we used Voigt's method (cf. original and modern works) based on a uniform strain. The calculated sets of elastic constants C ij (i and j corresponding to directions) are given in Table . All C ij values are positive. The elastic constants of qtz C 6 have the largest values, close to diamond , and smaller magnitudes were obtained for qtz BN and qtz SiC. The bulk (B V ) and shear (G V ) moduli (see the last two columns of Table ) were calculated using the equations for the hexagonal system : |
65153bcaa69febde9ed55d25 | 7 | Four modern theoretical models have been used to predict the Vickers hardness (H V ) of new phases. The thermodynamic (T) model , which is based on thermodynamic properties and crystal structure, generally shows good agreement with experiment, and is therefore recommended for hardness evaluation of superhard and ultrahard phases . The Vickers hardness and bulk modulus values calculated using this model are summarized in Table . The Lyakhov-Oganov (LO) model takes into account the topology of the crystal structure, the strength of covalent bonding, the degree of ionicity and directionality; and the empirical models, Mazhnik-Oganov (MO) and Chen-Niu (CN) , are based on elastic properties, namely, bulk and shear moduli. As shown previously , in the case of superhard (H V ≥ 40 GPa) compounds of light elements, the Lyakhov-Oganov model gives slightly underestimated values of hardness, while empirical models are not reliable. Fracture toughness (K Ic ) was evaluated using the Mazhnik-Oganov model . Table shows the hardness and other mechanical properties of the dense carbon, BN and SiC phases calculated using all four models. |
65153bcaa69febde9ed55d25 | 8 | The hardness and mechanical properties of qtz C 6 are, as expected, close to those of diamond and lonsdaleite. The corresponding values for qtz BN and especially for qtz SiC are significantly lower and are at the level of the mechanical properties of cubic boron nitride and cubic silicon carbide, respectively. It should be noted, however, that the hardness of all three phases with quartz topology is about 5% higher than that of the corresponding cubic phases. Such increased hardness is probably related to the ultrahigh densities of the qtz phases resulting from the distorted tetrahedron building blocks. |
65153bcaa69febde9ed55d25 | 9 | The bands develop along the main lines (horizontal direction) of the hexagonal Brillouin zone (reciprocal k-space). The vertical direction shows the frequencies ω, which are given in terahertz (THz). There are 3N phonon total modes with 3 acoustic modes starting from zero frequency (ω = 0 at the point, center of the Brillouin zone), up to a few terahertz and 3N-3 optical modes at frequencies higher than three. The three acoustic modes correspond to the lattice rigid translation modes with two transverse modes and one longitudinal mode. The remaining bands correspond to the optical modes. In all three subfigures there are no negative frequencies, and the corresponding carbon allotrope and two binary phases are dynamically stable. The latter indicates that these phases, once synthesized, can exist at ambient conditions. In qtz C 6 , the highest band culminates in the vicinity of ω ~ 40 THz, a value that has been observed for diamond by Raman spectroscopy . Binary compounds are characterized by lower energy bands. .... |
65153bcaa69febde9ed55d25 | 10 | The temperature dependencies of the heat capacity at constant volume (C v ) and entropy (S) of qtz С 6 , qtz BN and qtz SiC are shown in Fig. in comparison with experimental C v data for diamond , cubic BN and cubic SiC . It is quite expected that the heat capacities of all three phases with quartz topology formed by distorted tetrahedra are higher than the heat capacities of the corresponding cubic phases. |
65153bcaa69febde9ed55d25 | 11 | The bands develop along the main directions of the primitive hexagonal Brillouin zones. To the extent that all three phases exhibit band structures characterized by energy gaps between the valence band (VB) and the empty conduction band (CB), the energy reference along the vertical energy axis is with respect to the top of the VB: (E-E V ). qtz C 6 has a twice smaller band gap than diamond (~5 eV), showing a different behavior between dia and qtz topologies in the electronic structure behavior. The largest band gap is observed for qtz BN, which remains smaller than for cubic BN. The same feature of reduced band gap is observed for qtz SiC. It is also relevant to highlight that continuous VB in qtz C 6 versus two blocks in the binary compounds with a separation between low energy lying s-like states and higher energy lying p-like states up to E V . In conclusion, the phases with qtz topology are provided with enhanced covalence. |
65153bcaa69febde9ed55d25 | 12 | The band structure features are reflected in the site-projected electronic densities of states, DOS, shown in Fig. with energy along the horizontal axis and DOS in 1/eV units along the vertical axis. qtz C 6 , expressed as C1 3 C2 3 , is characterized by identical DOS for both sites and a continuous VB extending over ~27 eV, indicating the purely covalent nature of the carbon allotrope. The sharp DOS peak at 1 eV below EV belongs to the carbon p-states, which are more localized than the s-states smeared in the lower part of the VB. The (empty) CB also shows structured p-DOS. Turning to the binary compounds, the VB is now divided into two parts corresponding to s-states up to -15 eV in qtz BN (-9 eV in qtz SiC), followed by a broad block up to EV. The band gap in qtz BN is the largest, ~5 eV, which is close to c-BN, while qtz SiC has the smallest band gap of about 1.5 eV. |
65153bcaa69febde9ed55d25 | 13 | This paper presents a new class of binary compounds with quartz topology using boron nitride and silicon carbide as examples. The structures of qtz BN and qtz SiC have been constructed from the template carbon allotrope C 6 with quartz topology. It has been shown that the new phases are the densest among all known BN and SiC polymorphs. Accordingly, they are characterized by the highest hardness. In addition to mechanical stability, the new phases are also dynamically stable as indicated by the phonon band structures. The heat capacities of the new phases calculated from the phonon frequencies were found to be higher than those of the corresponding cubic phases; this is also true for qtz C 6 compared to diamond. It can be assumed that all of the above is a consequence of the presence of distorted tetrahedra in the crystal structures of the phases with quartz topology. Finally, from the analysis of electronic band structures and densities of states, it was found that the new phases exhibit semiconducting behavior. |
60cce9a026161123b48b1aaa | 0 | Nanotechnology has risen to prominence in the last decade as a potential game-changer in agricultural practices. Nanoscale particles have revolutionary properties that can increase pesticide efficiency and make the delivery system smart in terms of pesticide distribution. Pesticides, like nano-drugs, can be administered in a regulated and targeted manner using a smart delivery system. Nanoclays, such as montmorillonite (MMT), serve as nanofillers in polymer matrixes such as polysaccharides. Clay gallery swelling is induced by MMT alteration, which improves polysaccharide chain intercalation and clay dispersion in the polysaccharide gallery. Via modification and compounding with a polysaccharide, the silicate layers of MMT can be delaminated, resulting in a nanocomposite with improved tensile properties. When MNP is used instead of MMT as a nanofiller, the polysaccharide coating on MNP not only offers stability, but also reduces the toxicity of bare MNP, allowing the formulation to reach the target directly. The incorporation of nanofillers such as M MMT into the PEI results in a substantial improvement in the elasticity module, increased thermal resilience or greater fire retardancy factor. PEIs and macroscopic polymer such as, lignin greatly improve a range of physicochemical properties needed in new fields of application. (Figure ). Figure 1. Types of system evaluated for pesticide encapsulation and release In combination with nanotechnology, carbohydrates appear to be promising candidates for increasing E.E. We synthesized a sequence of hybrid M MMTs/MNPs of polysaccharides (chitosan, starch β-cyclodextrin CM-β-cyclodextrin lignin, PEIs forming hybrid nanocomposites by wet impregnation , gelation and coprecipitation in our attempt to further compare sustained pesticide release and enhanced E.E. methodologies (Figure ). |
60cce9a026161123b48b1aaa | 1 | Compared with M MMT PEI -M MMT nanocomposites (2,4,6,8,10 & 40 %), CS-M MMT S-M MMT, β-CD-M MMT nanocomposites and L-M MMT composites, the hybrid MNP NCs were designed as CS-MNP, CM-β-CD-MNP and compared with their EE % & PLE %age of chlorpyrifos. Finally, spectrophotometrically, sustained pesticide release (SPR) was determined wherein time-dependent release of chlorpyrifos from the above synthesized formulations was compared. (Figure ). Because of its phenol groups, which are capable of scavenging free radicals, Lignin was tried as it can serve as a stabilizer against UV degradation or thermo-oxidation (Figure ). Although the macro-to-nano-particles were discussed in a significant number of papers, it was shown that particle size reduction to nano shows pronounced E.E and SPR values due to more spaces (surface area) for insecticide molecules to be trapped. Another main aim of this research is to figure out which is the better nanotechnology (MNP vs M MMT). This is the first systematic relative account of E.E %, PLE wt % and polysaccharides vs. lignin vs. PEI aqueous release activity of chlorpyrifos using M MMT/MNP nanomaterials. Dialyzer tubes (MWCO 1KDa) were procured from G-Biosciences. Carboxymethyl-βcyclodextrin was synthesized from microcrystalline β-cyclodextrin as per literature procedure 36 2.2 Analytical methods. X-ray Diffraction (XRD) was performed using Bruker D-8 advanced diffractometer in the 2θ range of 10 to 90 °C. The average crystallite size with and without surface coating was estimated using the Scherrer equation. Fourier transform infrared (FT-IR) spectra were recorded on Agilent Cary 660 spectrometer using the KBr pellet technique in a range of 4000-400 cm -1 . Thermogravimetric analysis (TGA) was performed to determine the degradation/decomposition behavior of samples using thermogravimetric (TG) analyzer-(Perkin Elmer STA 8000) at a N2 flow rate of 10 mL/min and heating rate of 10 °C/min. Atomic Force Microscopy (AFM) images were acquired using Bruker Multimode 8 and sample analysis was done in tapping mode. The samples were deposited on silicon wafers and analysis was performed at different sections at room temperature and ambient atmosphere. |
60cce9a026161123b48b1aaa | 2 | Transmission Electron microscopy (TEM) images were acquired on a Jeol 2100 HR operating at 200 kV. Samples were prepared by depositing a drop of diluted NP suspension on 300 mesh TEM grid were dried under vacuum for 2 h. Inductively coupled plasma-mass spectrometer (ICP-MS) was used to estimate the amount of iron for calculating mol% and turn over frequency (TOF), Agilent Technologies 7700 series. |
60cce9a026161123b48b1aaa | 3 | The GC-MS analysis was performed to determine the encapsulation efficiency and pesticide loading efficiency using a Shimadzu GC coupled with a GCMS-QP 2010 plus mass detector and a single-quadrupole mass spectrometer Quantum (Shimadzu) with 100% dimethyl polysiloxane (Restek Rxi-1ms; 30 m × 0.25 mmiD., 0.25 μm film thickness) column. GC-MS operating conditions: The initial oven temperature was 60 °C, maintained for 1 min and then ramped to 270 °C at a rate of 10 °C/min followed by holding for 5 min at 270 °C. The initial temperature of the injector was 63 °C and then programmed at the same rate as the oven. |
60cce9a026161123b48b1aaa | 4 | Helium was used as carrier gas with primary pressure of 570 KPa. The split injection mode was used with a split ratio of 10.0. The injection volume of each sample was 1 μL and the total time for one GC-MS run was 27 min. Mass spectrometer settings: electron impact ionization mode with an electron energy of 70 eV, ion source and interface temperatures were set at 200 and 270 °C, respectively and scan mass range m/z 50-500. Shimadzu 2600 UV-Vis spectrophotometer was used to estimate the amount of entrapped pesticide at 214 nm during the aqueous release study after every 24 h. The calibration curve of ChP was prepared in MeOH. The synthesis of modified montmorillonite was adapted from the literature with minor changes to improve textural properties, especially pore volume and surface area of clay samples. 37 2g of clay was added to a 20ml solution of 6M hydrochloric acid. The mixture was then heated for 4 hours in an oil bath at 95°C while stirring at 600 rpm. The suspension was then washed with distilled water before being dried overnight in a 100°C oven. The powder was vacuumdried for another 24 hours before application. |
60cce9a026161123b48b1aaa | 5 | Approximately 2 g of PEI was dissolved in 20 g of methanol for 30 minutes while stirring. Wet impregnation method was used to make the PEI-M MMT composite sorbents, which was adapted from the literature with slight changes. The M MMT was then added to the abovementioned methanol solution and stirred at room temperature for 3h. The slurry was then dried for 12h at 75 0 C in a vacuum oven. PEI-M MMT samples with various PEI loadings of 2,4,6,8,10, and 40 wt percent were prepared to optimize the PEI loading on the M MMT. |
60cce9a026161123b48b1aaa | 6 | With minor modifications, the preparation of the CS-M MMT nanocomposite was adapted from the literature 1 g of Chitosan was mixed with 1 percent acetic acid to make Chitosan solutions (100 mL). After 1 hour of stirring at 60 °C, the solutions were stirred continuously at room temperature overnight. NaOH was used to get the pH of the solution down to 4.9. 5g of well-dispersed M MMT suspension was prepared in 100ml of water and held at 60 °C overnight with stirring, then centrifuged to remove any solids. M MMT dispersions (100 mL) were combined with chitosan solutions for 1 day at 60°C. The mixture was centrifuged for 5 min. |
60cce9a026161123b48b1aaa | 7 | With minor modifications, the β-cyclodextrin -M MMT solution was synthesized according to the literature. To produce a homogeneous mixture, β-cyclodextrin (1g) was dissolved in 25 ml acetic acid solution (1%) and stirred for 4 h. A suspension of 2% M MMT (500 mg) in acetic acid solution (1%, 25 ml) was also prepared. The suspension was added into the resulting gel. To produce a homogeneous β-CD-M MMT suspension, the mixture was stirred at 50°C for 2 days and then freeze dried. |
60cce9a026161123b48b1aaa | 8 | With minor modifications, CM-β-CD was synthesized as described in the literature. A solution of 16.3 percent monochloroacetic acid (2.7 ml) was used to treat a mixture of β-CD (1 g) and NaOH (0.93 g) in water (3.7 ml) at 50 0 C for 5 h. The pH values were modified in the range of 6-7 after the temperature of the reaction mixture was decreased to 25 °C. The obtained neutral solution was mixed with 10 mL of methanol to create a white carboxymethylated-β-cyclodextrin precipitate, which was purified and dried in a 50 °C oven. |
60cce9a026161123b48b1aaa | 9 | With minor modifications, CM-β-CD-MNP was synthesized according to published methods. With rapid stirring at a speed of 1200 rpm, 0.57 g FeCl2.4H2O, 1.57 g FeCl3.6H2O, and 1 g CM-β-CD were dissolved in 26.7 ml distilled water. As the reaction mixture reached 90 °C, 3.5 mL liquid ammonia (25%) was added in drops. The reaction was held at 90 °C for 1 hour with continuous stirring. The nanoparticles were then washed in distilled water to eliminate any unreacted contaminants before being dried in a 70 °C oven. |
60cce9a026161123b48b1aaa | 10 | The chitosan-coated MNP were made by in situ co-precipitation of iron salts, as described in the literature, but with a few tweaks 42 An amount of 3.6 × 10 -3 moles of iron from a mixture in a molar ratio 2 : 1 (Fe 3+ : Fe 2+ ) of ferric nitrate (3.19g) and ferrous sulfate (1.19g) added to chitosan (5g in 100ml) was mixed at 100 rpm in 3% (v/v) acetic acid(10ml) at 70 °C. An ultrasonic processor was used to spread the chitosan-iron solution, allowing for smoother compound delivery. Following that, a solution of 20% (w/v) NaOH : 96 percent (v/v) ethanol in a 12:3 volume ratio was added to the produced chitosan iron solution to precipitate it (30ml). The alkaline mixture was then homogenized using a vortex for 30 seconds before being shaken gently (60 rpm) for 18 h. After centrifugation for 5 minutes, the precipitate was washed in a 1:1 volume ratio of 50 mM phosphate buffer pH 7.0 and 96 percent (v/v) ethanol before neutralized. The neutralized solids were dried in an oven at 80 °C for 5 hours before being pounded to a fine powder. |
60cce9a026161123b48b1aaa | 11 | Minor modifications were made to the L-M MMT complex, as stated in the literature. 1 g M MMT and 1 g lignin were suspended in 30 mL sterile water and stirred for 12 h at room temperature. The L-M MMT complex was centrifuged and dried in a 65 °C oven for 24 h for powder processing. ). With the aid of an external magnet, the ChPloaded nanocarriers were collected, washed with methanol, and vacuum dried for 1 h at 37 °C. |
60cce9a026161123b48b1aaa | 12 | Samples of PEI-M MMT / Polysaccharide-M MMT/ Polysaccharide-MNPs/L-M MMT hybrid and inorganic (M MMT) of known weights (20 mg) were placed in a glass vial containing 15 ml of phosphate buffer at pH 6.5 at ambient temperature for slow release studies of loaded Chp loaded NCs. The set-up was gently shaken and the was weighed, pH was measured (Table in SI) and 1 ml of the liquid was removed for examination and supplemented with a fresh 1 ml of the medium to preserve the sink state. After syringe-filtering the aliquot, a UV-Vis spectrometer was used to investigate the release and correlate the concentration emitted with the Chp calibration map. |
60cce9a026161123b48b1aaa | 13 | We chose polysaccharides like Chitosan (CS), Starch (S), and β-Cyclodextrin (β-CD) for this analysis to compare polysaccharides like starch and cellulose NPs and MNPs described in our previous paper. The polysaccharide coating on MNP not only makes it more stable, but it also makes it less toxic, allowing the formulation to directly reach the target. We also contrasted it with M MMT, which works in the polymer matrix as nanofillers including polysaccharides, lignin, and PEI in the polymer matrix to significantly improve a range of physicochemical properties needed in new fields of application. As a result, in addition to comparing polysaccharides, lignin, and PEI, we also compare M MMT vs MNP encapsulation studies. |
60cce9a026161123b48b1aaa | 14 | The inner surface OH peaks at 3620 cm -1 and 3697 cm -1 persisted in the spectrum of the M MMT. The shift in the OH absorption band from (3441 cm -1 to 3443 cm -1 , 1627 cm -1 to 1626 cm -1 , and the absence of the 3410 cm -1 band suggested that modification was achieved on this site. Protons attack the -OH groups in the clay layers during acid activation of M MMT, creating changes in the adsorption bands ascribed to the OH vibration and octahedral cations. In CS-MNP, the band at 2,875 cm -1 was attributed to the symmetrical stretching of the -CH2 group in the polymer, 1,660 cm -1 to the amide I group (C-O stretching along the N-H deformation), 1,557 cm -1 to the -NH deformation, 1,412 cm -1 to the C-N axial deformation (amine group), 1,374 cm -1 to the COO-group in carboxylic acid salt, and 1,157 cm -1 to peak of β (1-4) glucosidic bond. |
60cce9a026161123b48b1aaa | 15 | The peak at 1704 cm -1 corresponds to carbonyl group (=CO) stretching, indicating that the carboxymethyl group (-COOCH3) has been incorporated into the β-CD molecule. With a minor shift, all of the prominent peaks of CM-β-CD in the region of 900-1200 cm -1 are present in the spectrum of CM-β-CD-MNP. two peaks formed at 1623 and 1401 cm -1 , indicating that the COOH groups of CM-β-CD reacted with the surface OH groups of Fe3O4 particles, resulting in the creation of the iron carboxylate. (Figure & S4 in SI). The digital morphology of the compounds was depicted in Figure & S6 in SI. 3.2g. DLS: DLS showed the size range for M MMTs/MNP, MNPs in the range of -228.9 to 3728 d-nm material absorption, material RI and viscosity of 0.00, 1.47 and 1.8872 respectively |
60cce9a026161123b48b1aaa | 16 | were quantified using gas chromatography-mass spectrometry (GC-MS) study and corrected with inductively coupled plasma mass spectrometry (ICP MS). (Table , Table , S4 in SI). These findings show that while slight alternation was found in terms of PLE, the PLE value was not greatly improved by functionalization. iii) CS-MNP outperforms β-CD-MNP among the hybrid NCs. Based on these findings, we infer that due to 1,4-& 1,6-linkages of the glucose unit, the high output of MNP over M MMT and better performance of S-M MMT may be attributed to the branching nature of starch, generating more space for pesticide packing. |
60cce9a026161123b48b1aaa | 17 | Although, Lignin-M MMT had good E.E/PLE results, but SPR is only 3.5% in 2days and therefore not suitable for aqueous release studies. In M MMT-PEIs, as we increase the organic content (PEI), the SPR results improve ( 40% M MMT-PEI: 81% in 10days while 2% M MMT-PEI: 65% in 11 days). S-M MMT showed significantly better performance(80.4wt%) in comparison to all other polysaccharides Among the M MMTs and hybrid MNP NCs, MNPs |
65673334cf8b3c3cd7652610 | 0 | Aziridines are a privileged motif within natural products and organic synthesis. This ring system is electrophilic enabling reaction with a broad range of nucleophiles and access to functionalized amine products. Aziridines have found extensive utility in the total synthesis of alkaloid natural products, as well as their downstream functionalization. Moreover, aziridine natural product analogues have demonstrated potent antitumor, antimicrobial, and antiviral activity. Such properties have made aziridines highly attractive structures for applications within the pharmaceutical industry, for example as covalent inhibitors. Considering their utility in medicinal chemistry, it is interesting to note a general lack of diversity of aziridine structures within the literature (Scheme 1a). Aliphatic and aromatic substituents dominate this space; however, few α-heterocyclic examples have been reported, with N-functionalized examples particularly scarce. This is perhaps due to the requirement of 'activated' nitrogen species (bearing electron-withdrawing groups) used in these methodologies, such as nitrenes. Despite extensive development, few examples of α-heterocyclic aziridines have been reported using transition metal catalysis, electrochemical or flow electrochemical approaches, organocatalysis, or utilising substitution chemistry with sulfur ylides or imines. Scheme 1. (a) Reported syntheses of α-heterocyclic aziridines. (b) Chiral Brønsted acid-catalyzed Michael addition/aziridination. (c) This work: Brønsted acid-catalyzed route to rare α-heterocyclic 1,2-discubstituted aziridines. |
65673334cf8b3c3cd7652610 | 1 | Previous work within our group sought to access this underexplored chemotype through a chiral phosphoric acid (CPA)-catalyzed enantioselective aza-Michael addition of aromatic amines to chlorovinyl heterocycles, followed by basic annulation (Scheme 1b). This yielded chiral mono-substituted α-heterocyclic aryl aziridines with high yield and ee. A limitation of this approach was the lack of further substitution on the aziridine ring. We therefore sought to expand the scope of this CPA strategy by establishing a method for the synthesis of 1,2disubsituted α-heterocyclic aryl aziridines. Here we provide a full account of this reaction development, including scope, limitations, and mechanism. |
65673334cf8b3c3cd7652610 | 2 | Following a preliminary assessment of reaction variables (see ESI for full details), the desired conjugate addition product 4 was obtained in 61% NMR yield and ca. 4:1 d.r. using chloroacetic acid as a catalyst (entry 1). Based on this, we explored the use of chiral phosphoric acid catalysts 5 and 6 to induce enantioselective bond formation (entries 2-4); however, despite similar d.r. to entry 1, levels of asymmetric induction were poor. Enantioselectivity could not be improved despite broad screening (see ESI). Despite an enantioselective process appearing unsuccessful, promising diastereoselectivity was observed using simple carboxylic acid catalysts (e.g., entry 1). We therefore sought to develop a diastereoselective approach to the rare disubstituted aziridine products. Performing the reaction in aromatic solvents, such as PhMe (entry 5) gave improved yields; however, at the expense of diastereocontrol. In contrast, excellent diastereoselectivity was obtained using polar protic solvents such as i-PrOH (entry 6), but at lower yield. Mixtures of i-PrOH and PhMe gave results in between the extremes of the pure solvents (entries 7 and 8). No further improvements to reaction conditions could be made. The significant impact of these solvents compelled further investigation (Scheme 3). process. Experiment and theory are in agreement when the barriers for the back reaction are considered. The reversions to 8 by the two stereoisomers have barriers of 25.3 and 26.9 kcal/mol; these are lower than the rate-limiting barrier leading to 8 and thus reversion of 4 and equilibration of the two isomers is to be expected. The initial emergence of a modest excess of anti-4 (see ESI for details) is consistent with this isomer having the higher barrier for reversion. The calculations support a thermodynamically controlled preference, and suggest any energy difference between the syn and anti isomers is below the limit of what can be accurately computed, supportive of the nonstereoselective formation observed. Turning to the stereoselective degradation of 4, it was assumed that this depends on trace Lewis or Brønsted acid that can complex the quinoline nitrogen of 4, hence the reaction proceeding in the presence of both i-PrOH and silica. Further computational studies used a concentration of 1 × 10 -18 M of protonated i-PrOH, 4, and settings for solvation by i-PrOH to model these conditions. Protonation of 4 promotes fragmentation with computed barriers of 26.6 kcal/mol for fragmentation of syn-4 and 28.1 kcal/mol for anti-4. This is consistent with the direction and degree of stereoselectivity observed and with catalysis by trace acid. The origin of the stereoselectivity can be understood by considering the effect of protonating the quinoline nitrogen. This creates a cationic aromatic surface (Figure ) that makes a strong interaction with the electron rich aniline ring, renedering the system pseudo-cyclic. A further stabilizing influence is a throughspace interaction between the electron-rich belt around the Cl and the nearby NH + . In the syn-11 structure (shown looking down the fragmenting bond in Figure ), it is clear that the aniline-π cation interaction and Cl-NH interaction can be comfortably retained while permitting the methyl group to be anti-periplanar to the quinoline. Inversion to give anti-11 might require loss of the Cl-NH interaction or the methyl clashing with the quinoline. Instead, the preferred structure retains these features but in doing so requires the pseudo-cyclic structure held together by the aniline-π cation interaction to be boat-like. Thus, syn-4 is more readily protonated and fragments more rapidly than anti-4. With no further optimization of the conjugate addition step possible, we proceeded with the high d.r. method and combined with a base-mediated SNi to enable access to the desired 1,2disubstituted aziridine targets (Scheme 4). |
65673334cf8b3c3cd7652610 | 3 | Anilines/amines In most cases this delivered the anti-chloroamine product as a single diastereomer. The anti-chloroamine could then be treated with LiHMDS to deliver the desired cis-aziridine with complete stereochemical fidelity (Scheme 5b). Regarding the generality of the two-step process, a range of heterocycles and aromatic amines were compatible. For the conjugate addition step (Scheme 5a), modification of the azaheterocycle was possible, with substituted quinoline cores achieving moderate to excellent yields (13-16). In general, more electron-rich systems, such as those bearing electron-donating groups (16) or benzothiazole (17) were less reactive. Under the PhMe-based conditions, product ratio does not vary greatly from 1.2:1; however, except for benzothiazole 17, the enrichment process generally delivered >20:1 d.r. A broad range of anilines was accommodated, with efficiency varying in line with electronic parameters. Electron-rich anilines (e.g., 2-OMe, 18) were generally more efficient, due to enhanced nucleophilicity. Conversely, anilines bearing electron-withdrawing groups (e.g., 2-CF3, 20) performed poorly for the same reason. The incorporation of synthetic handles, such as halogens (22, 23, 26) or Bpin (19) was also tolerated with good yields. Scalability of this first step was demonstrated in a 4 mmol reaction of the benchmark reaction to give 0.45 g of 4 as a single diastereomer. The chloroamine products underwent smooth conversion to the desired cis-aziridines in generally excellent yields. Confirmation of stereochemistry was obtained through single crystal X-ray structures of 12 and 45. Scalability was also demonstrated in the benchmark reaction to give 0.37 g of 12 as a single diastereomer. There are some limitations to the route outlined, as shown in Scheme 5c (see ESI for full details). Alternative heterocycles, such as pyridine, pyrimidine, or benzimidazole systems were unreactive. The reaction was highly sensitive to amine nucleophilicity, with more nucleophilic amines (e.g., alkylamines) resulting in catalyst deactivation. In addition, anilines with strong electron-withdrawing groups, such as nitrile or ester, were also unreactive. Derivatization of the alkene substituent was also challenging (Scheme 6). Changing from Me (3) to Ph (1), i-Pr (48), or c-Hex (50) was not accommodated. Increasing catalyst loading, temperature, aniline equivalents, or reaction time failed to enable the conjugate addition process using these chloroalkenes. Scheme 6. Substitution tolerance on the chloroalkene component. Yield determined by 1 H NMR using TCE as an internal standard. |
65673334cf8b3c3cd7652610 | 4 | To rationalize these observations, we once again turned to DFT calculations, using the solvation parameters for toluene. Having already established that the rate-limiting step is the initial addition, the reactants and transition states for the different substituted examples were obtained. These revealed that whereas the computed barrier for 3 is 27.3, the barriers for 1, 48, and 50 are 30.7, 27.5, and 29.2 kcal/mol, respectively. In general, a deactivating effect by larger groups is observed. Comparing the transition state for 3 (52) with that for 1 (53) reveals a significant difference between the two (Figure ). In this process, the double bond shifts along one carbon atom and in the transition state there are therefore two adjacent partial double bonds that would ideally have the three atoms involved and all of their substituents coplanar (substructure highlighted in blue in Figure ). For the reaction of 52, this can be tolerated and the C(Me)-C-C-Cl dihedral angle is 163°. However, co-planarity in the reaction of 1 would entail the Ph group clashing with the quinoline. To avoid this clash, the C(Ph)-C-C-Cl dihedral angle must reduce to 147° in transition state 53, and this species corresponds to a higher barrier than for 52. |
65673334cf8b3c3cd7652610 | 5 | In summary, an operationally simple synthesis of rare αheterocyclic aziridines has been developed via the Brønsted acidcatalyzed Michael addition of anilines to chloroalkenes. Degradation of the syn-diastereomer of the chloroamine products by trace acid allows the isolation of a single anti-diastereomer, which has been rationalized by computational modelling. The anti-chloroamine products can then be readily converted to the corresponding cis-aziridines with complete stereochemical fidelity. The scope of this transformation has been demonstrated, and despite some limitations, this approach offers a solution to a considerable gap in this area of chemical space that cannot currently be accessed by alternative synthetic methodologies. |
65673334cf8b3c3cd7652610 | 6 | General experimental procedure for preparing cisaziridines (e.g., 12) 3 (40.7 mg, 0.2 mmol, 1.0 equiv) and chloroacetic acid (5.7 mg, 0.06 mmol, 30 mol%) were added to a dry microwave vial, which was capped, purged, and filled with N2. PhMe (500 μL, 0.4 M) was added, and the mixture was stirred at 40 °C for 15 min prior to the addition of aniline (24 μL, 0.26 mmol, 1.3 equiv). The reaction was stirred at 40 °C for 16 h, before being quenched by the addition of saturated NaHCO3 solution. The mixture was extracted with EtOAc, dried over Na2SO4, filtered, and concentrated under reduced pressure. |
65673334cf8b3c3cd7652610 | 7 | The residue was resuspended in i-PrOH (1 mL, 0.2 M) and stirred at 40 °C for 4 h (or until a single diastereomer was present by 1 H NMR), and then concentrated under reduced pressure. The resulting product was suspended in dry THF (500 μL, 0.4 M) and stirred at RT for 2 min prior to the dropwise addition of LiHMDS (0.26 mmol, 1.3 equiv). The reaction was stirred at RT for 30 min, quenched with saturated NH4Cl solution, and extracted with EtOAc. The combined organic layers were dried over Na2SO4, filtered, and concentrated under reduced pressure. The crude product was purified by column chromatography (silica gel, 0-5% EtOAc in hexane) to yield the desired cis-aziridine 12 as a yellow oil (25 mg, 48%). Calculations employed the MN15 density functional and the 6-31+G** basis set. 15 Solvation was incorporated via the SMD continuum solvation model. All geometries were optimised and confirmed as minima or transition states by frequency calculation in Gaussian16. Free energies were computed with the goodvibes software with concentration of 1M, temperature of 298 K and a frequency cutoff of 20 cm -1 and the Grimme quasi-harmonic approach for low frequency vibrations. Systematic conformational sampling of all minima was undertaken with low energy conformations subjected to redundant internal coordinate scanning to obtain corresponding transition states. Electrostatic potential was computed in GaussView using default settings and molecular structure visualizations created in CYLview2. |
666bf9b7c9c6a5c07a7379e5 | 0 | Retrosynthetic analysis of Lenacapavir reveals four distinct building blocks 2, 3, 4 and highly functionalized central core 5a (Scheme 1). Efforts towards the synthesis of these building blocks chemically have already been reported. In view of the growing interest in the use of biocatalysis for sustainable manufacture of APIs, particularly those containing chiral amine building blocks, we envisaged preparing fragment 5a from the corresponding ketone 5 using an aminotransferase catalyzed reaction. Aminotransferases have emerged as a broadly useful class of biocatalyst for the preparation of both (S)-and (R)-configured amines but typically require extensive engineering to display good activity with sterically hindered ketones such as 5. Indeed, a survey of the literature to date revealed ketone 5 to be amongst the most sterically demanding and bulky candidate substrates for a transaminase catalyzed process. To identify an ATA with activity towards ketone 5, an in-house panel of >400 aminotransferases were initially evaluated using Lalanine as the amine donor coupled with an established pyruvate removal system. Unfortunately, from these assays we were unable to identify enzymes with detectable activity toward this sterically encumbered substrate. However, encouragingly several aminotransferases were identified with activity towards structurally related ketones (6 and 7) lacking one or both of the two bromine substituents on the pyridyl ring of the ketone substrate (Figure ). Interestingly no activity was observed with substrate 8 which contains a single 5-Br substituent ortho to the reacting ketone, highlighting this as a particularly challenging motif for aminotransferases. |
666bf9b7c9c6a5c07a7379e5 | 1 | Of the enzymes evaluated, ATA TA25 (previously reported as 3FCR-QM-W59L) gave the highest conversions of 6 and 7 to the respective amine products (Figure ). To unlock desired aminotransferase activity with ketone 5, we subjected TA25 to four rounds of directed evolution using a substrate walking approach (Figure and Supplementary Table ). In each round, ca. 20-24 positions were individually randomized using NNK degenerate codons. Individual library variants were arrayed in 96-well plates and evaluated as clarified cell lysate using UPLC assays. Beneficial mutations identified in each round were subsequently combined by DNA shuffling. |
666bf9b7c9c6a5c07a7379e5 | 2 | In rounds 1-2 variants were evaluated in the deamination direction using mono-brominated amine 8a as the substrate (note that the parent template had no activity towards 8a) (Figure and Supplementary Table ). A single F125G mutation installed during round 1 gave low levels of activity towards amine 8a, which was further improved in round 2. Pleasingly, the L59A mutation installed during round 2 also gave rise to our first activity towards deamination of the target di-brominated substrate 5a, which was then used as the substrate for the round 3 of engineering. For the fourth and final round of engineering, assays were performed in the synthesis direction using ketone 5 as the substrate. Here, isopropyl amine (iPA) was selected as the amine donor due to its low cost and previous implementation in industrial scale processes involving engineered aminotransferases. With a suitable aminotransferase in hand, we carried out a series of biotransformations to assess enzyme activity and selectivity under a range of substrate, enzyme and iPA concentrations. Using substrate 5 (2.5 mM) and iPA (500 mM) as the amine donor, our engineered aminotransferase affords the target product (S)-5 with 90% conversion and >99% e.e.) using 4 mol% enzyme (Figure and Table . At higher substrate loadings (5 mM), reasonable conversions of 70% and 50% can be achieved using 2 mol% and 1 mol% of enzyme, respectively. |
666bf9b7c9c6a5c07a7379e5 | 3 | In conclusion, we have engineered a highly selective aminotransferase for the synthesis of a key chiral intermediate of the anti-HIV drug Lenacapavir 1. The sterically demanding nature of the ketone substrate necessitated adoption of a 'substrate walk' approach to unlock desired activity starting from a parent template with no observable activity for the target transformation. Efforts are now ongoing to further engineer this enzyme into an industrial biocatalyst for the large-scale manufacture of Lenacapavir 1. |
666bf9b7c9c6a5c07a7379e5 | 4 | [ †] these authors contributed equally. Reverse phase HPLC and LCMS analysis was performed on a 1200 Series Agilent LC or LC/MSD system equipped with a G1379A degasser, G1313A binary pump, G1367A well plate auto sampler, G1316A temperature-controlled compartment and a diode array detector. For analysis an InfinityLab Poroshell 120 EC-C18, (4.6 mm x 100 mm, 4µm) LC Column was used. |
666bf9b7c9c6a5c07a7379e5 | 5 | Injection volumes of 2 μL were used and a DAD detector recorded absorbance at 210, 220 and 254nm. Substrates and products were eluted over 2 or 4 minutes using a gradient of 20 -95% acetonitrile (0.1% TFA) at 0.8 mL min -1 . Peaks were assigned by comparison to chemically synthesized standards or analysis of MS data and the peak areas were integrated using Agilent OpenLab software. |
666bf9b7c9c6a5c07a7379e5 | 6 | Normal-phase HPLC was carried out on an Agilent 1100 series system (USA) equipped with a G1379A degasser, G1313A binary pump, G1367A well plate auto sampler, G1316A temperature-controlled compartment and a diode array detector. For chiral analysis a Daicel (Osaka, Japan) CHIRALPAK® AD-H column (particle size 5 μm, dimensions: 4.6 x 250 mm) was used: 4.6 nm diameter and 250 mm length. Injection volumes of 10 μL were used and the DAD detector recorded absorbance at 254 nm. Substrates and products were eluted over 20 minutes using an isocratic method 10:90 (Hexane:IPA 0.1% DEA). Peaks were assigned by comparison to to chemically synthesized standards or analysis of MS data and the peak areas were integrated using Agilent OpenLab software. |
666bf9b7c9c6a5c07a7379e5 | 7 | The gene for TA25 (3FCR-QM-W59L) from Ruegeria sp. TM 1040 was codon optimised for expression in E. coli and synthesized and ligated into the pET28b vector by Twist Bioscience, between NdeI and XhoI restriction sites. Plasmid constructs were transformed into E. coli NEB5α and sequenced (MWG Eurofins) using T7 and T7term primers to verify the insert and then transformed into E. coli BL21 (DE3) for expression. |
666bf9b7c9c6a5c07a7379e5 | 8 | Overnight cultures were prepared by inoculating 10 mL of LB (supplemented with 50 μg mL -1 kanamycin) with a single colony and growing for 16 hours at 37 o C at 200 rpm. The overnight culture was then used to inoculate a baffled flask containing 400 mL TB media and 50 μg mL -1 kanamycin and incubated at 37 o C at 250 rpm until optical density at 600 nm (OD600) of 0.6. The culture was cooled to 20 o C and isopropyl-β-D-thiogalactoside (IPTG) was added to a final concentration of 0.1 mM to induce expression of the transaminase and further incubated overnight at 20 o C at 180 rpm. Cells were centrifuged at 4000 rpm for 30 min and stored as pellets at -20 o C. |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.