#!/usr/bin/env python3 """ Generate FASTA file with all protein sequences from previous step (for mmseqs2 clustering) Outputs: - A FASTA file of all proteins with UniProt sequence that was able to be pulled Usage: python 03_export_fasta.py \ --input ../data/fireprot_with_sequences.parquet \ --output ../data/proteins.fasta Notes: - This script has the option to include only residues which have consistent length of Uniprot sequence and sequence length specified in the original CSV. """ import argparse import pandas as pd def main(): ap = argparse.ArgumentParser() ap.add_argument("--input", default="../data/fireprotdb_with_sequences.parquet") ap.add_argument("--output_fasta", default="../data/proteins.fasta") ap.add_argument("--only_length_match", action="store_true") args = ap.parse_args() df = pd.read_parquet(args.input) # keep only rows with sequence df = df[df["sequence"].notna()].copy() if args.only_length_match and "length_match" in df.columns: df = df[df["length_match"] == True].copy() # one sequence per uniprotkb; fallback to sequence_id if needed df["uniprotkb"] = df["uniprotkb"].astype("string").fillna("").str.strip() df["sequence_id"] = df["sequence_id"].astype("string").fillna("").str.strip() # prefer uniprotkb as id, else sequence_id df["protein_id"] = df["uniprotkb"] df.loc[df["protein_id"] == "", "protein_id"] = "seqid:" + df.loc[df["protein_id"] == "", "sequence_id"] df.loc[df["protein_id"] == "seqid:", "protein_id"] = "unknown" # dedupe by protein_id (keep first) prot = df.drop_duplicates(subset=["protein_id"])[["protein_id", "sequence"]] with open(args.output_fasta, "w") as f: for _, r in prot.iterrows(): pid = r["protein_id"] seq = r["sequence"] if not isinstance(seq, str) or not seq: continue f.write(f">{pid}\n{seq}\n") print(f"Wrote {len(prot):,} proteins to {args.output_fasta}") if __name__ == "__main__": main()