protein-quest 0.3.0__py3-none-any.whl

protein_quest/cli.py

@@ -0,0 +1,782 @@
+"""Module for cli parsers and handlers."""
+
+import argparse
+import asyncio
+import csv
+import logging
+import os
+from collections.abc import Callable, Iterable
+from importlib.util import find_spec
+from io import TextIOWrapper
+from pathlib import Path
+from textwrap import dedent
+
+from cattrs import structure
+from rich import print as rprint
+from rich.logging import RichHandler
+from rich_argparse import ArgumentDefaultsRichHelpFormatter
+from tqdm.rich import tqdm
+
+from protein_quest.__version__ import __version__
+from protein_quest.alphafold.confidence import ConfidenceFilterQuery, filter_files_on_confidence
+from protein_quest.alphafold.fetch import DownloadableFormat, downloadable_formats
+from protein_quest.alphafold.fetch import fetch_many as af_fetch
+from protein_quest.emdb import fetch as emdb_fetch
+from protein_quest.filters import filter_files_on_chain, filter_files_on_residues
+from protein_quest.go import Aspect, allowed_aspects, search_gene_ontology_term, write_go_terms_to_csv
+from protein_quest.pdbe import fetch as pdbe_fetch
+from protein_quest.pdbe.io import glob_structure_files
+from protein_quest.taxonomy import SearchField, _write_taxonomy_csv, search_fields, search_taxon
+from protein_quest.uniprot import PdbResult, Query, search4af, search4emdb, search4pdb, search4uniprot
+
+logger = logging.getLogger(__name__)
+
+
+def _add_search_uniprot_parser(subparsers: argparse._SubParsersAction):
+    """Add search uniprot subcommand parser."""
+    parser = subparsers.add_parser(
+        "uniprot",
+        help="Search UniProt accessions",
+        description="Search for UniProt accessions based on various criteria in the Uniprot SPARQL endpoint.",
+        formatter_class=ArgumentDefaultsRichHelpFormatter,
+    )
+    parser.add_argument(
+        "output",
+        type=argparse.FileType("w", encoding="UTF-8"),
+        help="Output text file for UniProt accessions (one per line). Use `-` for stdout.",
+    )
+    parser.add_argument("--taxon-id", type=str, help="NCBI Taxon ID, e.g. 9606 for Homo Sapiens")
+    parser.add_argument(
+        "--reviewed",
+        action=argparse.BooleanOptionalAction,
+        help="Reviewed=swissprot, no-reviewed=trembl. Default is uniprot=swissprot+trembl.",
+        default=None,
+    )
+    parser.add_argument(
+        "--subcellular-location-uniprot",
+        type=str,
+        help="Subcellular location label as used by UniProt (e.g. nucleus)",
+    )
+    parser.add_argument(
+        "--subcellular-location-go",
+        dest="subcellular_location_go",
+        action="append",
+        help="GO term(s) for subcellular location (e.g. GO:0005634). Can be given multiple times.",
+    )
+    parser.add_argument(
+        "--molecular-function-go",
+        dest="molecular_function_go",
+        action="append",
+        help="GO term(s) for molecular function (e.g. GO:0003677). Can be given multiple times.",
+    )
+    parser.add_argument("--limit", type=int, default=10_000, help="Maximum number of uniprot accessions to return")
+    parser.add_argument("--timeout", type=int, default=1_800, help="Maximum seconds to wait for query to complete")
+
+
+def _add_search_pdbe_parser(subparsers: argparse._SubParsersAction):
+    """Add search pdbe subcommand parser."""
+    parser = subparsers.add_parser(
+        "pdbe",
+        help="Search PDBe structures of given UniProt accessions",
+        description="Search for PDB structures of given UniProt accessions in the Uniprot SPARQL endpoint.",
+        formatter_class=ArgumentDefaultsRichHelpFormatter,
+    )
+    parser.add_argument(
+        "uniprot_accs",
+        type=argparse.FileType("r", encoding="UTF-8"),
+        help="Text file with UniProt accessions (one per line). Use `-` for stdin.",
+    )
+    parser.add_argument(
+        "output_csv",
+        type=argparse.FileType("w", encoding="UTF-8"),
+        help=dedent("""\
+            Output CSV with `uniprot_acc`, `pdb_id`, `method`, `resolution`, `uniprot_chains`, `chain` columns.
+            Where `uniprot_chains` is the raw UniProt chain string, for example `A=1-100`.
+            and where `chain` is the first chain from `uniprot_chains`, for example `A`.
+            Use `-` for stdout.
+        """),
+    )
+    parser.add_argument(
+        "--limit", type=int, default=10_000, help="Maximum number of PDB uniprot accessions combinations to return"
+    )
+    parser.add_argument("--timeout", type=int, default=1_800, help="Maximum seconds to wait for query to complete")
+
+
+def _add_search_alphafold_parser(subparsers: argparse._SubParsersAction):
+    """Add search alphafold subcommand parser."""
+    parser = subparsers.add_parser(
+        "alphafold",
+        help="Search AlphaFold structures of given UniProt accessions",
+        description="Search for AlphaFold structures of given UniProt accessions in the Uniprot SPARQL endpoint.",
+        formatter_class=ArgumentDefaultsRichHelpFormatter,
+    )
+    parser.add_argument(
+        "uniprot_accs",
+        type=argparse.FileType("r", encoding="UTF-8"),
+        help="Text file with UniProt accessions (one per line). Use `-` for stdin.",
+    )
+    parser.add_argument(
+        "output_csv",
+        type=argparse.FileType("w", encoding="UTF-8"),
+        help="Output CSV with AlphaFold IDs per UniProt accession. Use `-` for stdout.",
+    )
+    parser.add_argument(
+        "--limit", type=int, default=10_000, help="Maximum number of Alphafold entry identifiers to return"
+    )
+    parser.add_argument("--timeout", type=int, default=1_800, help="Maximum seconds to wait for query to complete")
+
+
+def _add_search_emdb_parser(subparsers: argparse._SubParsersAction):
+    """Add search emdb subcommand parser."""
+    parser = subparsers.add_parser(
+        "emdb",
+        help="Search Electron Microscopy Data Bank (EMDB) identifiers of given UniProt accessions",
+        description=dedent("""\
+            Search for Electron Microscopy Data Bank (EMDB) identifiers of given UniProt accessions
+            in the Uniprot SPARQL endpoint.
+        """),
+        formatter_class=ArgumentDefaultsRichHelpFormatter,
+    )
+    parser.add_argument(
+        "uniprot_accs",
+        type=argparse.FileType("r", encoding="UTF-8"),
+        help="Text file with UniProt accessions (one per line). Use `-` for stdin.",
+    )
+    parser.add_argument(
+        "output_csv",
+        type=argparse.FileType("w", encoding="UTF-8"),
+        help="Output CSV with EMDB IDs per UniProt accession. Use `-` for stdout.",
+    )
+    parser.add_argument("--limit", type=int, default=10_000, help="Maximum number of EMDB entry identifiers to return")
+    parser.add_argument("--timeout", type=int, default=1_800, help="Maximum seconds to wait for query to complete")
+
+
+def _add_search_go_parser(subparsers: argparse._SubParsersAction):
+    """Add search go subcommand parser"""
+    parser = subparsers.add_parser(
+        "go",
+        help="Search for Gene Ontology (GO) terms",
+        description="Search for Gene Ontology (GO) terms in the EBI QuickGO API.",
+        formatter_class=ArgumentDefaultsRichHelpFormatter,
+    )
+    parser.add_argument(
+        "term",
+        type=str,
+        help="GO term to search for. For example `apoptosome`.",
+    )
+    parser.add_argument("--aspect", type=str, choices=allowed_aspects, help="Filter on aspect.")
+    parser.add_argument(
+        "output_csv",
+        type=argparse.FileType("w", encoding="UTF-8"),
+        help="Output CSV with GO term results. Use `-` for stdout.",
+    )
+    parser.add_argument("--limit", type=int, default=100, help="Maximum number of GO term results to return")
+
+
+def _add_search_taxonomy_parser(subparser: argparse._SubParsersAction):
+    """Add search taxonomy subcommand parser."""
+    parser = subparser.add_parser(
+        "taxonomy",
+        help="Search for taxon information in UniProt",
+        description=dedent("""\
+            Search for taxon information in UniProt.
+            Uses https://www.uniprot.org/taxonomy?query=*.
+        """),
+        formatter_class=ArgumentDefaultsRichHelpFormatter,
+    )
+    parser.add_argument(
+        "query", type=str, help="Search query for the taxon. Surround multiple words with quotes (' or \")."
+    )
+    parser.add_argument(
+        "output_csv",
+        type=argparse.FileType("w", encoding="UTF-8"),
+        help="Output CSV with taxonomy results. Use `-` for stdout.",
+    )
+    parser.add_argument(
+        "--field",
+        type=str,
+        choices=search_fields,
+        help=dedent("""\
+            Field to search in. If not given then searches all fields.
+            If "tax_id" then searches by taxon ID.
+            If "parent" then given a parent taxon ID returns all its children.
+            For example, if the parent taxon ID is 9606 (Human), it will return Neanderthal and Denisovan.
+        """),
+    )
+    parser.add_argument("--limit", type=int, default=100, help="Maximum number of results to return")
+
+
+def _add_retrieve_pdbe_parser(subparsers: argparse._SubParsersAction):
+    """Add retrieve pdbe subcommand parser."""
+    parser = subparsers.add_parser(
+        "pdbe",
+        help="Retrieve PDBe gzipped mmCIF files for PDB IDs in CSV.",
+        description=dedent("""\
+            Retrieve mmCIF files from Protein Data Bank in Europe Knowledge Base (PDBe) website
+            for unique PDB IDs listed in a CSV file.
+        """),
+        formatter_class=ArgumentDefaultsRichHelpFormatter,
+    )
+    parser.add_argument(
+        "pdbe_csv",
+        type=argparse.FileType("r", encoding="UTF-8"),
+        help="CSV file with `pdb_id` column. Other columns are ignored. Use `-` for stdin.",
+    )
+    parser.add_argument("output_dir", type=Path, help="Directory to store downloaded PDBe mmCIF files")
+    parser.add_argument(
+        "--max-parallel-downloads",
+        type=int,
+        default=5,
+        help="Maximum number of parallel downloads",
+    )
+
+
+def _add_retrieve_alphafold_parser(subparsers: argparse._SubParsersAction):
+    """Add retrieve alphafold subcommand parser."""
+    parser = subparsers.add_parser(
+        "alphafold",
+        help="Retrieve AlphaFold files for IDs in CSV",
+        description="Retrieve AlphaFold files from the AlphaFold Protein Structure Database.",
+        formatter_class=ArgumentDefaultsRichHelpFormatter,
+    )
+    parser.add_argument(
+        "alphafold_csv",
+        type=argparse.FileType("r", encoding="UTF-8"),
+        help="CSV file with `af_id` column. Other columns are ignored. Use `-` for stdin.",
+    )
+    parser.add_argument("output_dir", type=Path, help="Directory to store downloaded AlphaFold files")
+    parser.add_argument(
+        "--what-af-formats",
+        type=str,
+        action="append",
+        choices=sorted(downloadable_formats),
+        help=dedent("""AlphaFold formats to retrieve. Can be specified multiple times.
+            Default is 'pdb'. Summary is always downloaded as `<entryId>.json`."""),
+    )
+    parser.add_argument(
+        "--max-parallel-downloads",
+        type=int,
+        default=5,
+        help="Maximum number of parallel downloads",
+    )
+
+
+def _add_retrieve_emdb_parser(subparsers: argparse._SubParsersAction):
+    """Add retrieve emdb subcommand parser."""
+    parser = subparsers.add_parser(
+        "emdb",
+        help="Retrieve Electron Microscopy Data Bank (EMDB) gzipped 3D volume files for EMDB IDs in CSV.",
+        description=dedent("""\
+            Retrieve volume files from Electron Microscopy Data Bank (EMDB) website
+            for unique EMDB IDs listed in a CSV file.
+        """),
+        formatter_class=ArgumentDefaultsRichHelpFormatter,
+    )
+    parser.add_argument(
+        "emdb_csv",
+        type=argparse.FileType("r", encoding="UTF-8"),
+        help="CSV file with `emdb_id` column. Other columns are ignored. Use `-` for stdin.",
+    )
+    parser.add_argument("output_dir", type=Path, help="Directory to store downloaded EMDB volume files")
+
+
+def _add_filter_confidence_parser(subparsers: argparse._SubParsersAction):
+    """Add filter confidence subcommand parser."""
+    parser = subparsers.add_parser(
+        "confidence",
+        help="Filter AlphaFold mmcif/PDB files by confidence",
+        description=dedent("""\
+            Filter AlphaFold mmcif/PDB files by confidence (plDDT).
+            Passed files are written with residues below threshold removed."""),
+        formatter_class=ArgumentDefaultsRichHelpFormatter,
+    )
+    parser.add_argument("input_dir", type=Path, help="Directory with AlphaFold mmcif/PDB files")
+    parser.add_argument("output_dir", type=Path, help="Directory to write filtered mmcif/PDB files")
+    parser.add_argument("--confidence-threshold", type=float, default=70, help="pLDDT confidence threshold (0-100)")
+    parser.add_argument(
+        "--min-residues", type=int, default=0, help="Minimum number of high-confidence residues a structure should have"
+    )
+    parser.add_argument(
+        "--max-residues",
+        type=int,
+        default=10_000_000,
+        help="Maximum number of high-confidence residues a structure should have",
+    )
+    parser.add_argument(
+        "--write-stats",
+        type=argparse.FileType("w", encoding="UTF-8"),
+        help=dedent("""\
+            Write filter statistics to file.
+            In CSV format with `<input_file>,<residue_count>,<passed>,<output_file>` columns.
+            Use `-` for stdout."""),
+    )
+
+
+def _add_filter_chain_parser(subparsers: argparse._SubParsersAction):
+    """Add filter chain subcommand parser."""
+    parser = subparsers.add_parser(
+        "chain",
+        help="Filter on chain.",
+        description=dedent("""\
+            For each input PDB/mmCIF and chain combination
+            write a PDB/mmCIF file with just the given chain
+            and rename it to chain `A`.
+            Filtering is done in parallel using a Dask cluster."""),
+        formatter_class=ArgumentDefaultsRichHelpFormatter,
+    )
+    parser.add_argument(
+        "chains",
+        type=argparse.FileType("r", encoding="UTF-8"),
+        help="CSV file with `pdb_id` and `chain` columns. Other columns are ignored.",
+    )
+    parser.add_argument(
+        "input_dir",
+        type=Path,
+        help=dedent("""\
+        Directory with PDB/mmCIF files.
+        Expected filenames are `{pdb_id}.cif.gz`, `{pdb_id}.cif`, `{pdb_id}.pdb.gz` or `{pdb_id}.pdb`.
+    """),
+    )
+    parser.add_argument(
+        "output_dir",
+        type=Path,
+        help=dedent("""\
+        Directory to write the single-chain PDB/mmCIF files. Output files are in same format as input files."""),
+    )
+    parser.add_argument(
+        "--scheduler-address",
+        help="Address of the Dask scheduler to connect to. If not provided, will create a local cluster.",
+    )
+
+
+def _add_filter_residue_parser(subparsers: argparse._SubParsersAction):
+    """Add filter residue subcommand parser."""
+    parser = subparsers.add_parser(
+        "residue",
+        help="Filter PDB/mmCIF files by number of residues in chain A",
+        description=dedent("""\
+            Filter PDB/mmCIF files by number of residues in chain A.
+        """),
+        formatter_class=ArgumentDefaultsRichHelpFormatter,
+    )
+    parser.add_argument("input_dir", type=Path, help="Directory with PDB/mmCIF files (e.g., from 'filter chain')")
+    parser.add_argument(
+        "output_dir",
+        type=Path,
+        help=dedent("""\
+        Directory to write filtered PDB/mmCIF files. Files are copied without modification.
+    """),
+    )
+    parser.add_argument("--min-residues", type=int, default=0, help="Min residues in chain A")
+    parser.add_argument("--max-residues", type=int, default=10_000_000, help="Max residues in chain A")
+    parser.add_argument(
+        "--write-stats",
+        type=argparse.FileType("w", encoding="UTF-8"),
+        help=dedent("""\
+            Write filter statistics to file.
+            In CSV format with `<input_file>,<residue_count>,<passed>,<output_file>` columns.
+            Use `-` for stdout."""),
+    )
+
+
+def _add_search_subcommands(subparsers: argparse._SubParsersAction):
+    """Add search command and its subcommands."""
+    parser = subparsers.add_parser(
+        "search",
+        help="Search data sources",
+        description="Search various things online.",
+        formatter_class=ArgumentDefaultsRichHelpFormatter,
+    )
+    subsubparsers = parser.add_subparsers(dest="search_cmd", required=True)
+
+    _add_search_uniprot_parser(subsubparsers)
+    _add_search_pdbe_parser(subsubparsers)
+    _add_search_alphafold_parser(subsubparsers)
+    _add_search_emdb_parser(subsubparsers)
+    _add_search_go_parser(subsubparsers)
+    _add_search_taxonomy_parser(subsubparsers)
+
+
+def _add_retrieve_subcommands(subparsers: argparse._SubParsersAction):
+    """Add retrieve command and its subcommands."""
+    parser = subparsers.add_parser(
+        "retrieve",
+        help="Retrieve structure files",
+        description="Retrieve structure files from online resources.",
+        formatter_class=ArgumentDefaultsRichHelpFormatter,
+    )
+    subsubparsers = parser.add_subparsers(dest="retrieve_cmd", required=True)
+
+    _add_retrieve_pdbe_parser(subsubparsers)
+    _add_retrieve_alphafold_parser(subsubparsers)
+    _add_retrieve_emdb_parser(subsubparsers)
+
+
+def _add_filter_subcommands(subparsers: argparse._SubParsersAction):
+    """Add filter command and its subcommands."""
+    parser = subparsers.add_parser("filter", help="Filter files", formatter_class=ArgumentDefaultsRichHelpFormatter)
+    subsubparsers = parser.add_subparsers(dest="filter_cmd", required=True)
+
+    _add_filter_confidence_parser(subsubparsers)
+    _add_filter_chain_parser(subsubparsers)
+    _add_filter_residue_parser(subsubparsers)
+
+
+def _add_mcp_command(subparsers: argparse._SubParsersAction):
+    """Add MCP command."""
+
+    parser = subparsers.add_parser(
+        "mcp",
+        help="Run Model Context Protocol (MCP) server",
+        description=(
+            "Run Model Context Protocol (MCP) server. "
+            "Can be used by agentic LLMs like Claude Sonnet 4 as a set of tools."
+        ),
+        formatter_class=ArgumentDefaultsRichHelpFormatter,
+    )
+    parser.add_argument(
+        "--transport", default="stdio", choices=["stdio", "http", "streamable-http"], help="Transport protocol to use"
+    )
+    parser.add_argument("--host", default="127.0.0.1", help="Host to bind the server to")
+    parser.add_argument("--port", default=8000, type=int, help="Port to bind the server to")
+
+
+def make_parser() -> argparse.ArgumentParser:
+    parser = argparse.ArgumentParser(
+        description="Protein Quest CLI", prog="protein-quest", formatter_class=ArgumentDefaultsRichHelpFormatter
+    )
+    parser.add_argument("--log-level", default="WARNING", choices=["DEBUG", "INFO", "WARNING", "ERROR", "CRITICAL"])
+    parser.add_argument("--version", action="version", version=f"%(prog)s {__version__}")
+
+    subparsers = parser.add_subparsers(dest="command", required=True)
+
+    _add_search_subcommands(subparsers)
+    _add_retrieve_subcommands(subparsers)
+    _add_filter_subcommands(subparsers)
+    _add_mcp_command(subparsers)
+
+    return parser
+
+
+def main():
+    """Main entry point for the CLI."""
+    parser = make_parser()
+    args = parser.parse_args()
+    logging.basicConfig(level=args.log_level, handlers=[RichHandler(show_level=False)])
+
+    # Dispatch table to reduce complexity
+    cmd = args.command
+    sub = getattr(args, f"{cmd}_cmd", None)
+    handler = HANDLERS.get((cmd, sub))
+    if handler is None:
+        msg = f"Unknown command: {cmd} {sub}"
+        raise SystemExit(msg)
+    handler(args)
+
+
+def _handle_search_uniprot(args):
+    taxon_id = args.taxon_id
+    reviewed = args.reviewed
+    subcellular_location_uniprot = args.subcellular_location_uniprot
+    subcellular_location_go = args.subcellular_location_go
+    molecular_function_go = args.molecular_function_go
+    limit = args.limit
+    timeout = args.timeout
+    output_file = args.output
+
+    query = structure(
+        {
+            "taxon_id": taxon_id,
+            "reviewed": reviewed,
+            "subcellular_location_uniprot": subcellular_location_uniprot,
+            "subcellular_location_go": subcellular_location_go,
+            "molecular_function_go": molecular_function_go,
+        },
+        Query,
+    )
+    rprint("Searching for UniProt accessions")
+    accs = search4uniprot(query=query, limit=limit, timeout=timeout)
+    rprint(f"Found {len(accs)} UniProt accessions, written to {output_file.name}")
+    _write_lines(output_file, sorted(accs))
+
+
+def _handle_search_pdbe(args):
+    uniprot_accs = args.uniprot_accs
+    limit = args.limit
+    timeout = args.timeout
+    output_csv = args.output_csv
+
+    accs = set(_read_lines(uniprot_accs))
+    rprint(f"Finding PDB entries for {len(accs)} uniprot accessions")
+    results = search4pdb(accs, limit=limit, timeout=timeout)
+    total_pdbs = sum([len(v) for v in results.values()])
+    rprint(f"Found {total_pdbs} PDB entries for {len(results)} uniprot accessions")
+    rprint(f"Written to {output_csv.name}")
+    _write_pdbe_csv(output_csv, results)
+
+
+def _handle_search_alphafold(args):
+    uniprot_accs = args.uniprot_accs
+    limit = args.limit
+    timeout = args.timeout
+    output_csv = args.output_csv
+
+    accs = _read_lines(uniprot_accs)
+    rprint(f"Finding AlphaFold entries for {len(accs)} uniprot accessions")
+    results = search4af(accs, limit=limit, timeout=timeout)
+    rprint(f"Found {len(results)} AlphaFold entries, written to {output_csv.name}")
+    _write_dict_of_sets2csv(output_csv, results, "af_id")
+
+
+def _handle_search_emdb(args):
+    uniprot_accs = args.uniprot_accs
+    limit = args.limit
+    timeout = args.timeout
+    output_csv = args.output_csv
+
+    accs = _read_lines(uniprot_accs)
+    rprint(f"Finding EMDB entries for {len(accs)} uniprot accessions")
+    results = search4emdb(accs, limit=limit, timeout=timeout)
+    total_emdbs = sum([len(v) for v in results.values()])
+    rprint(f"Found {total_emdbs} EMDB entries, written to {output_csv.name}")
+    _write_dict_of_sets2csv(output_csv, results, "emdb_id")
+
+
+def _handle_search_go(args):
+    term = structure(args.term, str)
+    aspect: Aspect | None = args.aspect
+    limit = structure(args.limit, int)
+    output_csv: TextIOWrapper = args.output_csv
+
+    if aspect:
+        rprint(f"Searching for GO terms matching '{term}' with aspect '{aspect}'")
+    else:
+        rprint(f"Searching for GO terms matching '{term}'")
+    results = asyncio.run(search_gene_ontology_term(term, aspect=aspect, limit=limit))
+    rprint(f"Found {len(results)} GO terms, written to {output_csv.name}")
+    write_go_terms_to_csv(results, output_csv)
+
+
+def _handle_search_taxonomy(args):
+    query: str = args.query
+    field: SearchField | None = args.field
+    limit: int = args.limit
+    output_csv: TextIOWrapper = args.output_csv
+
+    if field:
+        rprint(f"Searching for taxon information matching '{query}' in field '{field}'")
+    else:
+        rprint(f"Searching for taxon information matching '{query}'")
+    results = asyncio.run(search_taxon(query=query, field=field, limit=limit))
+    rprint(f"Found {len(results)} taxons, written to {output_csv.name}")
+    _write_taxonomy_csv(results, output_csv)
+
+
+def _handle_retrieve_pdbe(args):
+    pdbe_csv = args.pdbe_csv
+    output_dir = args.output_dir
+    max_parallel_downloads = args.max_parallel_downloads
+
+    pdb_ids = _read_column_from_csv(pdbe_csv, "pdb_id")
+    rprint(f"Retrieving {len(pdb_ids)} PDBe entries")
+    result = asyncio.run(pdbe_fetch.fetch(pdb_ids, output_dir, max_parallel_downloads=max_parallel_downloads))
+    rprint(f"Retrieved {len(result)} PDBe entries")
+
+
+def _handle_retrieve_alphafold(args):
+    download_dir = args.output_dir
+    what_af_formats = args.what_af_formats
+    alphafold_csv = args.alphafold_csv
+    max_parallel_downloads = args.max_parallel_downloads
+
+    if what_af_formats is None:
+        what_af_formats = {"pdb"}
+
+    # TODO besides `uniprot_acc,af_id\n` csv also allow headless single column format
+    #
+    af_ids = [r["af_id"] for r in _read_alphafold_csv(alphafold_csv)]
+    validated_what: set[DownloadableFormat] = structure(what_af_formats, set[DownloadableFormat])
+    rprint(f"Retrieving {len(af_ids)} AlphaFold entries with formats {validated_what}")
+    afs = af_fetch(af_ids, download_dir, what=validated_what, max_parallel_downloads=max_parallel_downloads)
+    total_nr_files = sum(af.nr_of_files() for af in afs)
+    rprint(f"Retrieved {total_nr_files} AlphaFold files and {len(afs)} summaries, written to {download_dir}")
+
+
+def _handle_retrieve_emdb(args):
+    emdb_csv = args.emdb_csv
+    output_dir = args.output_dir
+
+    emdb_ids = _read_column_from_csv(emdb_csv, "emdb_id")
+    rprint(f"Retrieving {len(emdb_ids)} EMDB entries")
+    result = asyncio.run(emdb_fetch(emdb_ids, output_dir))
+    rprint(f"Retrieved {len(result)} EMDB entries")
+
+
+def _handle_filter_confidence(args: argparse.Namespace):
+    # we are repeating types here and in add_argument call
+    # TODO replace argparse with modern alternative like cyclopts
+    # to get rid of duplication
+    input_dir = structure(args.input_dir, Path)
+    output_dir = structure(args.output_dir, Path)
+    confidence_threshold = structure(args.confidence_threshold, float)
+    # TODO add min/max
+    min_residues = structure(args.min_residues, int)
+    max_residues = structure(args.max_residues, int)
+    stats_file: TextIOWrapper | None = args.write_stats
+
+    output_dir.mkdir(parents=True, exist_ok=True)
+    input_files = sorted(glob_structure_files(input_dir))
+    nr_input_files = len(input_files)
+    rprint(f"Starting confidence filtering of {nr_input_files} mmcif/PDB files in {input_dir} directory.")
+    query = structure(
+        {
+            "confidence": confidence_threshold,
+            "min_threshold": min_residues,
+            "max_threshold": max_residues,
+        },
+        ConfidenceFilterQuery,
+    )
+    if stats_file:
+        writer = csv.writer(stats_file)
+        writer.writerow(["input_file", "residue_count", "passed", "output_file"])
+
+    passed_count = 0
+    for r in tqdm(filter_files_on_confidence(input_files, query, output_dir), total=len(input_files), unit="file"):
+        if r.filtered_file:
+            passed_count += 1
+        if stats_file:
+            writer.writerow([r.input_file, r.count, r.filtered_file is not None, r.filtered_file])
+
+    rprint(f"Filtered {passed_count} mmcif/PDB files by confidence, written to {output_dir} directory")
+    if stats_file:
+        rprint(f"Statistics written to {stats_file.name}")
+
+
+def _handle_filter_chain(args):
+    input_dir = args.input_dir
+    output_dir = args.output_dir
+    pdb_id2chain_mapping_file = args.chains
+    scheduler_address = args.scheduler_address
+
+    rows = list(_iter_csv_rows(pdb_id2chain_mapping_file))
+    id2chains: dict[str, str] = {row["pdb_id"]: row["chain"] for row in rows}
+
+    new_files = filter_files_on_chain(input_dir, id2chains, output_dir, scheduler_address)
+
+    nr_written = len([r for r in new_files if r[2] is not None])
+
+    rprint(f"Wrote {nr_written} single-chain PDB/mmCIF files to {output_dir}.")
+
+
+def _handle_filter_residue(args):
+    input_dir = structure(args.input_dir, Path)
+    output_dir = structure(args.output_dir, Path)
+    min_residues = structure(args.min_residues, int)
+    max_residues = structure(args.max_residues, int)
+    stats_file: TextIOWrapper | None = args.write_stats
+
+    if stats_file:
+        writer = csv.writer(stats_file)
+        writer.writerow(["input_file", "residue_count", "passed", "output_file"])
+
+    nr_passed = 0
+    input_files = sorted(glob_structure_files(input_dir))
+    nr_total = len(input_files)
+    rprint(f"Filtering {nr_total} files in {input_dir} directory by number of residues in chain A.")
+    for r in filter_files_on_residues(input_files, output_dir, min_residues=min_residues, max_residues=max_residues):
+        if stats_file:
+            writer.writerow([r.input_file, r.residue_count, r.passed, r.output_file])
+        if r.passed:
+            nr_passed += 1
+
+    rprint(f"Wrote {nr_passed} files to {output_dir} directory.")
+    if stats_file:
+        rprint(f"Statistics written to {stats_file.name}")
+
+
+def _handle_mcp(args):
+    if find_spec("fastmcp") is None:
+        msg = "Unable to start MCP server, please install `protein-quest[mcp]`."
+        raise ImportError(msg)
+
+    from protein_quest.mcp_server import mcp  # noqa: PLC0415
+
+    if args.transport == "stdio":
+        mcp.run(transport=args.transport)
+    else:
+        mcp.run(transport=args.transport, host=args.host, port=args.port)
+
+
+HANDLERS: dict[tuple[str, str | None], Callable] = {
+    ("search", "uniprot"): _handle_search_uniprot,
+    ("search", "pdbe"): _handle_search_pdbe,
+    ("search", "alphafold"): _handle_search_alphafold,
+    ("search", "emdb"): _handle_search_emdb,
+    ("search", "go"): _handle_search_go,
+    ("search", "taxonomy"): _handle_search_taxonomy,
+    ("retrieve", "pdbe"): _handle_retrieve_pdbe,
+    ("retrieve", "alphafold"): _handle_retrieve_alphafold,
+    ("retrieve", "emdb"): _handle_retrieve_emdb,
+    ("filter", "confidence"): _handle_filter_confidence,
+    ("filter", "chain"): _handle_filter_chain,
+    ("filter", "residue"): _handle_filter_residue,
+    ("mcp", None): _handle_mcp,
+}
+
+
+def _read_lines(file: TextIOWrapper) -> list[str]:
+    return [line.strip() for line in file]
+
+
+def _make_sure_parent_exists(file: TextIOWrapper):
+    if file.name != "<stdout>":
+        Path(file.name).parent.mkdir(parents=True, exist_ok=True)
+
+
+def _write_lines(file: TextIOWrapper, lines: Iterable[str]):
+    _make_sure_parent_exists(file)
+    file.writelines(line + os.linesep for line in lines)
+
+
+def _write_pdbe_csv(path: TextIOWrapper, data: dict[str, set[PdbResult]]):
+    _make_sure_parent_exists(path)
+    fieldnames = ["uniprot_acc", "pdb_id", "method", "resolution", "uniprot_chains", "chain"]
+    writer = csv.DictWriter(path, fieldnames=fieldnames)
+    writer.writeheader()
+    for uniprot_acc, entries in sorted(data.items()):
+        for e in sorted(entries, key=lambda x: (x.id, x.method)):
+            writer.writerow(
+                {
+                    "uniprot_acc": uniprot_acc,
+                    "pdb_id": e.id,
+                    "method": e.method,
+                    "resolution": e.resolution or "",
+                    "uniprot_chains": e.uniprot_chains,
+                    "chain": e.chain,
+                }
+            )
+
+
+def _write_dict_of_sets2csv(file: TextIOWrapper, data: dict[str, set[str]], ref_id_field: str):
+    _make_sure_parent_exists(file)
+    fieldnames = ["uniprot_acc", ref_id_field]
+
+    writer = csv.DictWriter(file, fieldnames=fieldnames)
+    writer.writeheader()
+    for uniprot_acc, ref_ids in sorted(data.items()):
+        for ref_id in sorted(ref_ids):
+            writer.writerow({"uniprot_acc": uniprot_acc, ref_id_field: ref_id})
+
+
+def _read_alphafold_csv(file: TextIOWrapper):
+    reader = csv.DictReader(file)
+    yield from reader
+
+
+def _iter_csv_rows(file: TextIOWrapper):
+    reader = csv.DictReader(file)
+    yield from reader
+
+
+def _read_column_from_csv(file: TextIOWrapper, column: str) -> set[str]:
+    return {row[column] for row in _iter_csv_rows(file)}