import requests import tensorflow as tf import pandas as pd import numpy as np from operator import add from functools import reduce from Bio import SeqIO from Bio.SeqRecord import SeqRecord from Bio.SeqFeature import SeqFeature, FeatureLocation from Bio.Seq import Seq from keras.models import load_model import random import pyBigWig # configure GPUs for gpu in tf.config.list_physical_devices('GPU'): tf.config.experimental.set_memory_growth(gpu, enable=True) if len(tf.config.list_physical_devices('GPU')) > 0: tf.config.experimental.set_visible_devices(tf.config.list_physical_devices('GPU')[0], 'GPU') ntmap = {'A': (1, 0, 0, 0), 'C': (0, 1, 0, 0), 'G': (0, 0, 1, 0), 'T': (0, 0, 0, 1) } def get_seqcode(seq): return np.array(reduce(add, map(lambda c: ntmap[c], seq.upper()))).reshape( (1, len(seq), -1)) from keras.models import load_model class DCModelOntar: def __init__(self, ontar_model_dir, is_reg=False): self.model = load_model(ontar_model_dir) def ontar_predict(self, x, channel_first=True): if channel_first: x = x.transpose([0, 2, 3, 1]) yp = self.model.predict(x) return yp.ravel() def fetch_ensembl_transcripts(gene_symbol): url = f"https://rest.ensembl.org/lookup/symbol/homo_sapiens/{gene_symbol}?expand=1;content-type=application/json" response = requests.get(url) if response.status_code == 200: gene_data = response.json() if 'Transcript' in gene_data: return gene_data['Transcript'] else: print("No transcripts found for gene:", gene_symbol) return None else: print(f"Error fetching gene data from Ensembl: {response.text}") return None def fetch_ensembl_sequence(transcript_id): url = f"https://rest.ensembl.org/sequence/id/{transcript_id}?content-type=application/json" response = requests.get(url) if response.status_code == 200: sequence_data = response.json() if 'seq' in sequence_data: return sequence_data['seq'] else: print("No sequence found for transcript:", transcript_id) return None else: print(f"Error fetching sequence data from Ensembl: {response.text}") return None def find_crispr_targets(sequence, chr, start, strand, transcript_id, exon_id, pam="NGG", target_length=20): targets = [] len_sequence = len(sequence) complement = {'A': 'T', 'T': 'A', 'C': 'G', 'G': 'C'} dnatorna = {'A': 'A', 'T': 'U', 'C': 'C', 'G': 'G'} if strand == -1: sequence = ''.join([complement[base] for base in sequence]) for i in range(len_sequence - len(pam) + 1): if sequence[i + 1:i + 3] == pam[1:]: if i >= target_length: target_seq = sequence[i - target_length:i + 3] tar_start = start + i - target_length tar_end = start + i + 3 gRNA = ''.join([dnatorna[base] for base in sequence[i - target_length:i]]) targets.append([target_seq, gRNA, chr, str(tar_start), str(tar_end), str(strand), transcript_id, exon_id]) return targets # Function to predict on-target efficiency and format output def format_prediction_output(targets, model_path): dcModel = DCModelOntar(model_path) formatted_data = [] for target in targets: # Encode the gRNA sequence encoded_seq = get_seqcode(target[0]).reshape(-1,4,1,23) # Predict on-target efficiency using the model prediction = dcModel.ontar_predict(encoded_seq) # Format output gRNA = target[1] chr = target[2] start = target[3] end = target[4] strand = target[5] transcript_id = target[6] exon_id = target[7] formatted_data.append([chr, start, end, strand, transcript_id, exon_id, target[0], gRNA, prediction[0]]) return formatted_data def process_gene(gene_symbol, model_path): transcripts = fetch_ensembl_transcripts(gene_symbol) results = [] all_exons = [] # To accumulate all exons all_gene_sequences = [] # To accumulate all gene sequences if transcripts: for transcript in transcripts: Exons = transcript['Exon'] all_exons.extend(Exons) # Add all exons from this transcript to the list transcript_id = transcript['id'] for exon in Exons: exon_id = exon['id'] gene_sequence = fetch_ensembl_sequence(exon_id) if gene_sequence: all_gene_sequences.append(gene_sequence) # Add this gene sequence to the list start = exon['start'] strand = exon['strand'] chr = exon['seq_region_name'] targets = find_crispr_targets(gene_sequence, chr, start, strand, transcript_id, exon_id) if targets: # Predict on-target efficiency for each gRNA site formatted_data = format_prediction_output(targets, model_path) results.extend(formatted_data) else: print(f"Failed to retrieve gene sequence for exon {exon_id}.") else: print("Failed to retrieve transcripts.") # Return the sorted output, combined gene sequences, and all exons return results, all_gene_sequences, all_exons def create_genbank_features(df): features = [] for index, row in df.iterrows(): # Convert strand from '+/-' to 1/-1 for Biopython strand = 1 if row['Strand'] == '+' else -1 if row['Strand'] == '-' else 0 # Create feature location using the 'Start Pos' and 'End Pos' location = FeatureLocation(start=int(row['Start Pos']), end=int(row['End Pos']), strand=strand) # Create a SeqFeature feature = SeqFeature(location=location, type="misc_feature", qualifiers={ 'label': row['gRNA'], # Use gRNA as the label 'target': row['Target'], # Include the target sequence 'note': f"Prediction: {row['Prediction']:.4f}", # Include the prediction score 'transcript_id': row['Transcript'], 'exon_id': row['Exon'] }) features.append(feature) return features def generate_genbank_file_from_df(df, gene_sequence, gene_symbol, output_path): features = create_genbank_features(df) record = SeqRecord(Seq(gene_sequence), id=gene_symbol, name=gene_symbol, description=f'CRISPR Cas9 predicted targets for {gene_symbol}', features=features) record.annotations["molecule_type"] = "DNA" SeqIO.write(record, output_path, "genbank") def create_bed_file_from_df(df, output_path): with open(output_path, 'w') as bed_file: for index, row in df.iterrows(): chrom = row["Chr"] start = int(row["Start Pos"]) # Assuming 'Start Pos' is the column name in the df end = int(row["End Pos"]) # Assuming 'End Pos' is the column name in the df strand = '+' if row["Strand"] == '1' else '-' # Assuming 'Strand' is the column name in the df gRNA = row["gRNA"] score = str(row["Prediction"]) transcript_id = row["Transcript"] # Assuming 'Transcript' is the column name in the df bed_file.write(f"{chrom}\t{start}\t{end}\t{gRNA}\t{score}\t{strand}\t{transcript_id}\n") def create_csv_from_df(df, output_path): df.to_csv(output_path, index=False) def create_bigwig(df, bigwig_path, chrom_sizes_path): bw = pyBigWig.open(bigwig_path, "w") # Load chromosome sizes chrom_sizes = {} with open(chrom_sizes_path, 'r') as f: for line in f: chrom, size = line.strip().split() chrom_sizes[chrom] = int(size) bw.addHeader(list(chrom_sizes.items())) for chrom in df['Chr'].unique(): chrom_df = df[df['Chr'] == chrom] bw.addEntries( chrom, chrom_df['Start Pos'].astype(int).tolist(), ends=chrom_df['End Pos'].astype(int).tolist(), values=chrom_df['Prediction'].astype(float).tolist() ) bw.close()