myFunctions.py

import os
import sys
import streamlit as st
import pandas as pd
import joblib
import time
import numpy as np
from rdkit import Chem
from rdkit.Chem import Draw
from rdkit.Chem import AllChem
from rdkit import RDLogger
import uuid
from datasets import load_dataset
import requests
from io import BytesIO
import urllib.request
import networkx as nx
import warnings
warnings.filterwarnings('ignore')
from model import OnTheFlyModel, HitSelectorByOverlap, CommunityDetector, task_evaluator
from morgan_desc import *
from physchem_desc import *
from fragment_embedder import FragmentEmbedder

SIMILARITY_PERCENTILES = [95, 90]

def get_session_id():
    if "session_id" not in st.session_state:
        st.session_state["session_id"] = str(uuid.uuid4())
    return st.session_state["session_id"]

def clear_old_cache(cache_folder, hours=24):
    # Define the directory to check
    folder_path = cache_folder

    # Get the current time
    current_time = time.time()

    # Loop through each file in the directory
    for filename in os.listdir(folder_path):
        file_path = os.path.join(folder_path, filename)
        # Check if it's a file (and not a directory)
        if os.path.isfile(file_path):
            # Get the last modified time and compare
            file_modified_time = os.path.getmtime(file_path)
            if current_time - file_modified_time > hours*3600: # 3600 seconds in an hour
                print(f"Deleting {filename} as it is older than one day")
                os.remove(file_path) # Delete the file

def load_protein_spearman_similarity_matrix():
#    uniprot_acs, M = joblib.load(os.path.join(root, "..", "data", "protein_protein_spearman_correlations.joblib"))   
    url = ''.join(('https://huggingface.co/datasets/ligdis/data/resolve/main/protein_protein_spearman_correlations.joblib')) # The URL of the file you want to load
    with urllib.request.urlopen(url) as response:     # Download the file
        uniprot_acs, M = joblib.load(BytesIO(response.read()))
    values = np.triu(M, k=1).ravel()
    cutoffs = [np.percentile(values, p) for p in SIMILARITY_PERCENTILES]
    return uniprot_acs, M, cutoffs

def load_protein_hit_similarity_matrix():
#    uniprot_acs, M = joblib.load(os.path.join(root, "..", "data", "protein_protein_hit_cosines.joblib"))
    url = ''.join(('https://huggingface.co/datasets/ligdis/data/resolve/main/protein_protein_hit_cosines.joblib')) # The URL of the file you want to load
    with urllib.request.urlopen(url) as response:     # Download the file
        uniprot_acs, M = joblib.load(BytesIO(response.read()))
    values = np.triu(M, k=1).ravel()
    cutoffs = [np.percentile(values, p) for p in SIMILARITY_PERCENTILES]
    return uniprot_acs, M, cutoffs

# print("Loading uniprot acs")
global_uniprot_acs_0, M0, cutoffs_0 = load_protein_spearman_similarity_matrix()
global_uniprot_acs_1, M1, cutoffs_1 = load_protein_hit_similarity_matrix()
# print("Done")

def get_protein_graph(uniprot_acs):
    G = nx.Graph()
    G.add_nodes_from(uniprot_acs)
    pid2idx_0 = dict((k, i) for i, k in enumerate(global_uniprot_acs_0))
    pid2idx_1 = dict((k, i) for i, k in enumerate(global_uniprot_acs_1))
    for i, pid_0 in enumerate(uniprot_acs):
        for j, pid_1 in enumerate(uniprot_acs):
            if i >= j:
                continue
            v = M0[pid2idx_0[pid_0], pid2idx_0[pid_1]]
            for cutoff in cutoffs_0:
                if v >= cutoff:
                    if not G.has_edge(pid_0, pid_1):
                        G.add_edge(pid_0, pid_1, weight=1)
                    else:
                        current_weight = G[pid_0][pid_1].get("weight")
                        G[pid_0][pid_1]["weight"] = current_weight + 1
            v = M1[pid2idx_1[pid_0], pid2idx_1[pid_1]]
            for cutoff in cutoffs_1:
                if v >= cutoff:
                    if not G.has_edge(pid_0, pid_1):
                        G.add_edge(pid_0, pid_1, weight=1)
                    else:
                        current_weight = G[pid_0][pid_1].get("weight")
                        G[pid_0][pid_1]["weight"] = current_weight + 1
    return G

def load_hits():
#    hits, fid_prom, pid_prom = joblib.load(os.path.join(root, "..", "data", "hits.joblib"))
    url = ''.join(('https://huggingface.co/datasets/ligdis/data/resolve/main/hits.joblib')) # The URL of the file you want to load
    with urllib.request.urlopen(url) as response:     # Download the file
        hits, fid_prom, pid_prom = joblib.load(BytesIO(response.read()))
    return hits, fid_prom, pid_prom

def load_fid2smi():
#    d = pd.read_csv(os.path.join(root, "..", "data", "cemm_smiles.csv"))
    dataset = load_dataset('ligdis/data', data_files={"cemm_smiles.csv"})
    d = dataset['train'].to_pandas() 
    fid2smi = {}
    for r in d.values:
        fid2smi[r[0]] = r[1]
    return fid2smi

def pid2name_mapper():
#    df = pd.read_csv(os.path.join(root, "..", "data/pid2name_primary.tsv"), sep="\t", header=None)
    dataset = load_dataset('ligdis/data', data_files={"pid2name_primary.tsv"})
    df = dataset['train'].to_pandas()  
       
    df.columns = ["uniprot_ac", "gene_name"]
    name2pid = {}
    pid2name = {}
    any2pid = {}
    for r in df.values:
        name2pid[r[1]] = r[0]
        pid2name[r[0]] = r[1]
        any2pid[r[0]] = r[0]
        any2pid[r[1]] = r[0]
    return pid2name, name2pid, any2pid

def pids_to_dataframe(pids, pid2name, pid_prom):
    R = []
    for pid in pids:
        r = [pid, pid2name[pid], pid_prom[pid]]
        R += [r]
    df = (
        pd.DataFrame(R, columns=["UniprotAC", "Gene Name", "Fragment Hits"])
        .drop_duplicates()
        .reset_index(drop=True)
    )
    return df

def is_valid_smiles(smiles):
    try:
        mol = Chem.MolFromSmiles(smiles)
    except:
        mol = None
    if mol is None:
        return False
    else:
        return True

def has_crf(mol, CRF_PATTERN):
    pattern = CRF_PATTERN
    has_pattern = mol.HasSubstructMatch(Chem.MolFromSmarts(pattern))
    if not has_pattern:
        if mol.HasSubstructMatch(
            Chem.MolFromSmarts(CRF_PATTERN_0)
        ) and mol.HasSubstructMatch(Chem.MolFromSmarts(CRF_PATTERN_1)):
            return True
        else:
            return False
    return True

def attach_crf(smiles):
    mol = Chem.MolFromSmiles(smiles)
    combined_mol_0 = CombineMols(mol, crf_0, "O")
    combined_mol_1 = []
    combined_mol = combined_mol_0 + combined_mol_1
    result = []
    for cm in combined_mol:
        smi = Chem.MolToSmiles(cm)
        if "." in smi:
            continue
        mol = Chem.MolFromSmiles(smi)
        if mol is None:
            continue
        if not has_crf(mol):
            continue
        result += [Chem.MolToSmiles(mol)]
    if len(result) > 0:
        return result[0]
    else:
        return None

def get_fragment_image(smiles):
    m = Chem.MolFromSmiles(smiles)
    AllChem.Compute2DCoords(m)
    im = Draw.MolToImage(m, size=(200, 200))
    return im