promptadmin/drug-discovery-prompts
1
0
Fork 0
Code Issues Pull Requests 28 Packages Projects Releases Wiki Activity
Files
upstream-sync/scientific-agent-skills-20260818-9e8b0c-fcfk
drug-discovery-prompts/upstream/mims-harvard-ToolUniverse/skills/tooluniverse-infectious-disease/TOOLS_REFERENCE.md
T

15 KiB
Raw Permalink Blame History

title, task, lineage_type, upstream_source, upstream_sha, imported_at, prompt_class, upstream_changes, author, validated
title task lineage_type upstream_source upstream_sha imported_at prompt_class upstream_changes author validated
Infectious Disease Outbreak Intelligence - Tool Reference import https://github.com/mims-harvard/ToolUniverse/blob/e2520a96/skills/tooluniverse-infectious-disease/TOOLS_REFERENCE.md e2520a96 2026-06-26 prompt accepted upstream false

Infectious Disease Outbreak Intelligence - Tool Reference

Phase 1: Pathogen Identification

NCBI Taxonomy Tools

Tool Purpose Key Parameters
NCBIDatasets_suggest_taxonomy Search taxonomy database query
NCBIDatasets_get_taxonomy Get details by TaxID taxid
NCBIDatasets_get_taxonomy Get full lineage taxid

Example - Classify pathogen:

# Search for pathogen
tax = tu.tools.NCBIDatasets_suggest_taxonomy(query="SARS-CoV-2")
# Returns: {"taxid": 2697049, "scientific_name": "...", "lineage": [...]}

UniProt Protein Tools

Tool Purpose Key Parameters
UniProt_search Search proteins query, organism
UniProt_get_entry_by_accession Get protein details accession
UniProt_get_sequence_by_accession Get sequence accession

Example - Get viral proteins:

# Search for viral proteins
proteins = tu.tools.UniProt_search(
    query="organism:2697049",  # SARS-CoV-2 TaxID
    reviewed=True
)

Phase 2: Target Identification

ChEMBL Target Tools

Tool Purpose Key Parameters
ChEMBL_search_targets Search targets query, organism
ChEMBL_get_target_activities Get bioactivity target_chembl_id
ChEMBL_search_drugs Search drugs query, max_phase

Example - Find drug precedent:

# Search for protease inhibitors
drugs = tu.tools.ChEMBL_search_drugs(
    query="main protease coronavirus",
    max_phase=4  # Approved drugs only
)

DGIdb Tools

Tool Purpose Key Parameters
DGIdb_get_drug_gene_interactions Drug-target interactions genes
DGIdb_get_gene_druggability Druggability score genes

Phase 3: Structure Prediction (NVIDIA NIM)

Structure Prediction Tools

Tool Purpose Key Parameters Async
NvidiaNIM_alphafold2 High-accuracy prediction sequence, algorithm Yes
NvidiaNIM_esmfold Fast prediction sequence No
NvidiaNIM_openfold2 Alternative predictor sequence Yes

Example - Predict target structure:

# High-accuracy prediction
structure = tu.tools.NvidiaNIM_alphafold2(
    sequence=protease_sequence,
    algorithm="mmseqs2",
    relax_prediction=False
)
# Returns: {"structure": "<PDB content>", "plddt": [...]}

Structure Validation

def assess_structure_quality(structure_result):
    """Assess structure quality for docking."""
    plddt = structure_result.get('plddt', [])
    
    mean_plddt = np.mean(plddt)
    high_conf = sum(1 for p in plddt if p > 90) / len(plddt)
    
    return {
        'mean_plddt': mean_plddt,
        'high_confidence_fraction': high_conf,
        'docking_suitable': mean_plddt > 70 and high_conf > 0.5
    }

Phase 4: Drug Repurposing

ChEMBL Drug Search

Tool Purpose Key Parameters
ChEMBL_search_drugs Search approved drugs query, max_phase
ChEMBL_get_molecule Get drug details molecule_chembl_id
ChEMBL_get_drug_mechanisms Get MOA molecule_chembl_id

DrugBank Tools

Tool Purpose Key Parameters
drugbank_vocab_search Search drugs query
drugbank_get_drug_basic_info_by_drug_name_or_id Get drug details drugbank_id
drugbank_get_targets_by_drug_name_or_drugbank_id Get drug targets drugbank_id

Docking Tools (NVIDIA NIM)

Tool Purpose Key Parameters
NvidiaNIM_diffdock Blind docking protein, ligand, num_poses
NvidiaNIM_boltz2 Complex prediction polymers, ligands

Example - Dock drug candidates:

# Dock drug against target
result = tu.tools.NvidiaNIM_diffdock(
    protein=target_pdb_content,
    ligand=drug_smiles,
    num_poses=10
)
# Returns: {"poses": [{"confidence": 0.94, "coordinates": ...}, ...]}

Phase 4.5: Pathway Analysis (NEW)

KEGG Pathway Tools

Tool Purpose Key Parameters
kegg_search_pathway Search pathways query
KEGG_get_pathway_genes Get genes in pathway pathway_id
kegg_get_gene_info Get gene details gene_id
kegg_find_genes Find genes by keyword query, database

Example - Pathogen metabolism pathways:

# Search for viral replication pathways
pathways = tu.tools.kegg_search_pathway(
    query="coronavirus replication"
)

# Get essential genes
genes = tu.tools.KEGG_get_pathway_genes(
    pathway_id="ko03030"  # DNA replication
)

Reactome Tools

Tool Purpose Key Parameters
ReactomeContent_search Search pathways query, species
Reactome_get_participants Get pathway entities pathway_id
Reactome_get_pathway_hierarchy Get pathway tree pathway_id

Example - Host-pathogen interaction pathways:

# Host response to infection
pathways = tu.tools.ReactomeContent_search(
    query="viral infection response",
    species="Homo sapiens"
)

Phase 5: Literature Intelligence (ENHANCED)

PubMed Tools

Tool Purpose Key Parameters
PubMed_search_articles Search articles query, limit
PubMed_get_article Get article pmid

Example - Search outbreak literature:

papers = tu.tools.PubMed_search_articles(
    query="SARS-CoV-2 treatment drug",
    limit=50,
    sort="date"
)

Preprint Servers (CRITICAL for Outbreaks)

Tool Purpose Key Parameters
EuropePMC_search_articles Search preprints (bioRxiv, medRxiv) query, source='PPR', pageSize
ArXiv_search_papers Physics/ML preprints query, category
BioRxiv_get_preprint Get preprint by DOI doi, server='biorxiv'
MedRxiv_get_preprint Get preprint by DOI doi, server='medrxiv'

⚠️ Preprints are NOT peer-reviewed but critical for emerging outbreaks!

Example - Search preprints (bioRxiv/medRxiv don't have search APIs, use EuropePMC):

# Search for newest preprint findings
preprints = tu.tools.EuropePMC_search_articles(
    query=f"{pathogen_name} mechanism resistance",
    source="PPR",  # PPR = Preprints (bioRxiv, medRxiv, etc.)
    pageSize=20
)

# If you have a specific DOI, retrieve full metadata:
if doi_from_search.startswith('10.1101/'):
    full_preprint = tu.tools.BioRxiv_get_preprint(doi=doi_from_search)

# Alternative: Use web search for bioRxiv
web_results = tu.tools.web_search(
    query=f"{pathogen_name} clinical trial effectiveness",
    limit=20
)

# Computational papers
arxiv = tu.tools.ArXiv_search_papers(
    query=f"{pathogen_name} drug discovery",
    category="q-bio",
    limit=10
)

Citation Analysis Tools

Tool Purpose Key Parameters
openalex_search_works Search with citations query, limit
SemanticScholar_search_papers AI-ranked search query, limit

Example - Find high-impact papers:

# Get citation counts
papers = tu.tools.openalex_search_works(
    query="remdesivir COVID-19 trial",
    limit=20
)
# Returns: {"cited_by_count": 5234, ...}

# AI-ranked papers
ranked = tu.tools.SemanticScholar_search_papers(
    query="SARS-CoV-2 drug resistance",
    limit=20
)

Clinical Trials Tools

Tool Purpose Key Parameters
search_clinical_trials Search trials condition, intervention, status
ClinicalTrials_get_study Get trial details nct_id

Example - Find active trials:

trials = tu.tools.search_clinical_trials(
    condition="COVID-19",
    intervention="antiviral",
    status="Recruiting"
)

Workflow Code Examples

Example 1: Complete Outbreak Analysis

def analyze_outbreak(tu, pathogen_name):
    """Complete outbreak intelligence workflow."""
    
    # Phase 1: Identify pathogen
    taxonomy = tu.tools.NCBIDatasets_suggest_taxonomy(query=pathogen_name)
    taxid = taxonomy['taxid']
    
    # Phase 2: Get target proteins
    proteins = tu.tools.UniProt_search(
        query=f"organism:{taxid}",
        reviewed=True
    )
    
    # Phase 3: Predict structures for top targets
    structures = {}
    for protein in proteins[:3]:  # Top 3 targets
        seq = tu.tools.UniProt_get_sequence_by_accession(
            accession=protein['accession']
        )
        struct = tu.tools.NvidiaNIM_alphafold2(sequence=seq)
        structures[protein['name']] = struct
    
    # Phase 4: Find repurposing candidates
    candidates = tu.tools.ChEMBL_search_drugs(
        query=f"{pathogen_name} OR broad spectrum antiviral",
        max_phase=4
    )
    
    # Dock top candidates
    docking_results = []
    for drug in candidates[:20]:
        result = tu.tools.NvidiaNIM_diffdock(
            protein=structures['main_protease']['structure'],
            ligand=drug['smiles'],
            num_poses=5
        )
        docking_results.append({
            'drug': drug,
            'score': result['poses'][0]['confidence']
        })
    
    # Phase 5: Literature search
    papers = tu.tools.PubMed_search_articles(
        query=f"{pathogen_name} treatment",
        limit=50
    )
    
    return {
        'taxonomy': taxonomy,
        'targets': proteins,
        'structures': structures,
        'drug_candidates': sorted(docking_results, 
                                  key=lambda x: x['score'], 
                                  reverse=True),
        'literature': papers
    }

Example 2: Rapid Drug Screen

def rapid_drug_screen(tu, target_sequence, drug_smiles_list):
    """Rapid docking screen for drug repurposing."""
    
    # Quick structure prediction
    structure = tu.tools.NvidiaNIM_esmfold(sequence=target_sequence)
    
    # Dock all candidates
    results = []
    for smiles in drug_smiles_list:
        docking = tu.tools.NvidiaNIM_diffdock(
            protein=structure['structure'],
            ligand=smiles,
            num_poses=3
        )
        results.append({
            'smiles': smiles,
            'score': docking['poses'][0]['confidence']
        })
    
    return sorted(results, key=lambda x: x['score'], reverse=True)

Example 3: Knowledge Transfer from Related Pathogen

def transfer_knowledge(tu, novel_pathogen, reference_pathogen):
    """Transfer drug knowledge from related pathogen."""
    
    # Get drugs approved for reference pathogen
    ref_drugs = tu.tools.ChEMBL_search_drugs(
        query=reference_pathogen,
        max_phase=4
    )
    
    # Get target from novel pathogen
    novel_proteins = tu.tools.UniProt_search(
        query=f"organism:{novel_pathogen}"
    )
    
    # Find homologous targets
    homologs = []
    for protein in novel_proteins:
        # BLAST against reference
        blast = tu.tools.BLAST_protein_search(
            sequence=protein['sequence'],
            database="refseq_protein",
            organism=reference_pathogen
        )
        if blast and blast[0]['identity'] > 70:
            homologs.append({
                'novel_target': protein,
                'reference_homolog': blast[0],
                'identity': blast[0]['identity']
            })
    
    # Match drugs to homologous targets
    candidates = []
    for drug in ref_drugs:
        for homolog in homologs:
            if drug['target'] == homolog['reference_homolog']['accession']:
                candidates.append({
                    'drug': drug,
                    'target_homology': homolog['identity'],
                    'expected_activity': 'High' if homolog['identity'] > 90 else 'Medium'
                })
    
    return candidates

Fallback Chains

Taxonomy

Primary Fallback 1 Fallback 2
NCBIDatasets_suggest_taxonomy UniProtTaxonomy_search Manual NCBI query

Structure Prediction

Primary Fallback 1 Fallback 2
NvidiaNIM_alphafold2 NvidiaNIM_esmfold alphafold_get_prediction
alphafold_get_prediction NvidiaNIM_openfold2 PDB homolog

Docking

Primary Fallback 1 Fallback 2
NvidiaNIM_diffdock NvidiaNIM_boltz2 Literature docking

Drug Search

Primary Fallback 1 Fallback 2
ChEMBL_search_drugs drugbank_vocab_search PubChem BioAssay

Pathway Analysis (NEW)

Primary Fallback 1 Fallback 2
kegg_search_pathway ReactomeContent_search WikiPathways_search
KEGG_get_pathway_genes Reactome_get_participants Gene list extraction

Literature (ENHANCED)

Primary Fallback 1 Fallback 2
PubMed_search_articles openalex_search_works Google Scholar
EuropePMC_search_articles (source='PPR') web_search (site:biorxiv.org) ArXiv q-bio
openalex_search_works SemanticScholar_search_papers Manual citation

Common Parameter Mistakes

Tool Wrong Correct
NCBIDatasets_suggest_taxonomy name="virus" query="virus"
UniProt_search name="protease" query="protease"
ChEMBL_search_targets target="Mpro" query="Mpro"
NvidiaNIM_diffdock protein_file=path protein=content
NvidiaNIM_alphafold2 seq="MVLS..." sequence="MVLS..."

NVIDIA NIM Requirements

API Key: NVIDIA_API_KEY environment variable required

Rate limits: 40 RPM (1.5 second minimum between calls)

Async operations:

  • AlphaFold2 may return 202, requiring polling
  • ESMFold is synchronous (faster for rapid screening)

Check Availability

import os

nvidia_available = bool(os.environ.get("NVIDIA_API_KEY"))
if not nvidia_available:
    print("Warning: NVIDIA NIM tools unavailable, using fallbacks")

Speed Optimization

For Urgent Outbreaks

  1. Use ESMFold first for rapid structure (30 sec vs 5-15 min)
  2. Dock FDA-approved only initially (fastest to deploy)
  3. Parallelize docking if possible
  4. Cache structures for repeated queries

Prioritization Order

def prioritize_candidates(candidates):
    """Prioritize by speed to clinical use."""
    return sorted(candidates, key=lambda x: (
        -x['fda_approved'],      # FDA approved first
        -x['phase'],             # Higher phase next
        -x['docking_score']      # Then by score
    ))
Reference in New Issue View Git Blame Copy Permalink