import streamlit as st from stmol import showmol import py3Dmol import requests import biotite.structure.io as bsio import random import hashlib import urllib3 from Bio.Blast import NCBIWWW, NCBIXML from Bio.Seq import Seq from Bio.SeqRecord import SeqRecord import time urllib3.disable_warnings(urllib3.exceptions.InsecureRequestWarning) st.set_page_config(layout='wide') st.sidebar.title('🔮 GenPro2 Protein Generator & Structure Predictor') st.sidebar.write('GenPro2 is an end-to-end single sequence protein generator and structure predictor based [*ESMFold*](https://esmatlas.com/about) and the ESM-2 language model.') # Function to generate protein sequence from random words def generate_sequence_from_words(words, length): seed = ' '.join(words).encode('utf-8') random.seed(hashlib.md5(seed).hexdigest()) amino_acids = "ACDEFGHIKLMNPQRSTVWY" return ''.join(random.choice(amino_acids) for _ in range(length)) # stmol def render_mol(pdb): pdbview = py3Dmol.view() pdbview.addModel(pdb,'pdb') pdbview.setStyle({'cartoon':{'color':'spectrum'}}) pdbview.setBackgroundColor('white') pdbview.zoomTo() pdbview.zoom(2, 800) pdbview.spin(True) showmol(pdbview, height = 500,width=800) # BLAST analysis function def perform_blast_analysis(sequence): st.subheader('BLAST Analysis') with st.spinner("Analyzing generated protein... This may take a few minutes."): progress_bar = st.progress(0) for i in range(100): progress_bar.progress(i + 1) if i == 99: # Simulate longer process at the end time.sleep(2) try: record = SeqRecord(Seq(sequence), id='random_protein') result_handle = NCBIWWW.qblast("blastp", "swissprot", record.seq) blast_record = NCBIXML.read(result_handle) st.write('Top BLAST Match:') if blast_record.alignments: alignment = blast_record.alignments[0] # Get the top hit hsp = alignment.hsps[0] # Get the first (best) HSP # Extract protein name and organism title_parts = alignment.title.split('|') protein_name = title_parts[-1].strip() organism = title_parts[-2].split('OS=')[-1].split('OX=')[0].strip() # Calculate identity percentage identity_percentage = (hsp.identities / alignment.length) * 100 st.write(f"**Protein:** {protein_name}") st.write(f"**Organism:** {organism}") st.write(f"**Sequence Identity:** {identity_percentage:.2f}%") # Fetch protein function (if available) if hasattr(alignment, 'description') and alignment.description: st.write(f"**Possible Function:** {alignment.description}") # Link to BLAST blast_link = f"https://blast.ncbi.nlm.nih.gov/Blast.cgi?PROGRAM=blastp&PAGE_TYPE=BlastSearch&LINK_LOC=blasthome" st.markdown(f"[View full BLAST results]({blast_link})") else: st.write("No significant matches found.") except Exception as e: st.error(f"An error occurred during BLAST analysis: {str(e)}") st.write("Please try again later or contact support if the issue persists.") # ESMfold def update(sequence, word1, word2, word3, sequence_length): headers = { 'Content-Type': 'application/x-www-form-urlencoded', } try: response = requests.post('https://api.esmatlas.com/foldSequence/v1/pdb/', headers=headers, data=sequence, verify=False, # Disable SSL verification timeout=300) # Set a longer timeout response.raise_for_status() # Raise an exception for bad status codes pdb_string = response.content.decode('utf-8') with open('predicted.pdb', 'w') as f: f.write(pdb_string) struct = bsio.load_structure('predicted.pdb', extra_fields=["b_factor"]) b_value = round(struct.b_factor.mean(), 2) # Display protein structure st.subheader(f'Predicted protein structure using seed: {word1}, {word2}, and {word3} + length {sequence_length}') render_mol(pdb_string) # plDDT value is stored in the B-factor field st.subheader('plDDT Score') st.write('plDDT is a per-residue estimate of the confidence in prediction on a scale from 0-100%.') st.info(f'Average plDDT: {int(b_value * 100)}%') st.download_button( label="Download PDB", data=pdb_string, file_name='predicted.pdb', mime='text/plain', ) # Perform BLAST analysis perform_blast_analysis(sequence) except requests.exceptions.RequestException as e: st.error(f"An error occurred while calling the API: {str(e)}") st.write("Please try again later or contact support if the issue persists.") # Streamlit app st.title("Word-Seeded Protein Sequence Generator and Structure Predictor") # Input for word-seeded sequence generation st.sidebar.subheader("Generate Sequence from Words") word1 = st.sidebar.text_input("Word 1") word2 = st.sidebar.text_input("Word 2") word3 = st.sidebar.text_input("Word 3") sequence_length = st.sidebar.number_input("Sequence Length", min_value=50, max_value=400, value=100, step=10) # Generate and predict button if st.sidebar.button('Generate and Predict'): if word1 and word2 and word3: sequence = generate_sequence_from_words([word1, word2, word3], sequence_length) st.sidebar.text_area("Generated Sequence", sequence, height=100) st.sidebar.info("Note: The same words and sequence length will always produce the same sequence.") with st.spinner("Predicting protein structure... This may take a few minutes."): update(sequence, word1, word2, word3, sequence_length) else: st.sidebar.warning("Please enter all three words to generate a sequence.") # Information display st.sidebar.markdown(""" ## What to do next: If you find interesting results from the sequence folding, you can explore further: 1. Learn more about protein structures and sequences. 2. Visit the [Protein Data Bank (PDB)](https://www.rcsb.org/) for known protein structures. 3. Compare your folded structure with known functional proteins by downloading your results. 4. Read about similar proteins to gain insights into potential functions. **Remember, this folding is based on randomly generated sequences. Interpret the results with caution. Enjoy exploring the world of protein sequences! Share your high-confidence protein images with us on X [*@WandsAI*](https://x.com/wandsai)! """)