Life Science Database Query

Life-Science Database Query Skill

This skill packages a broad set of modular sub-skills that can be composed to answer questions across every major area of life-sciences research. Each sub-skill maps to one or more public databases and carries its own operating rules and execution scripts.

When To Use This Skill

• The user asks a broad life-sciences question ("what is known about …", "tell me about this gene/variant/compound …")
• The user wants a specific public-database lookup (gnomAD, UniProt, ChEMBL, ClinicalTrials.gov, etc.)
• The user needs multi-source evidence synthesis (genetics + expression, structure + chemistry, clinical + literature)
• The user wants locus-to-gene prioritization, PheWAS follow-up, or multi-cohort replication
• The user wants to discover public datasets or preprints for a research topic

Entry Point

For broad or ambiguous requests, invoke research-router-skill first. It classifies the request, normalizes entities, selects the minimum useful set of downstream sub-skills, optionally parallelizes independent evidence lanes using subagents, and synthesizes the final answer.

Use a specific sub-skill directly only when the request is clearly scoped to a single source.

Sub-Skill Directory

Sub-skill files live at skills//SKILL.md. Read the relevant SKILL.md before invoking any sub-skill.

Orchestration

Human Genetics And Variant Evidence

Expression, Cell Context, And Functional Genomics

Protein, Structure, Pathway, And Functional Biology

Chemistry, Metabolites, And Pharmacology

Clinical, Translational, And Disease Evidence

Literature, Search, And Public Study Discovery

Multi-Omics, Proteomics, And Specialized Data Sources

Routing Heuristics

Choose the minimum sub-skills needed to answer the question.

Subagent Guidance

When subagents are available, use them as bounded retrieval and analysis accelerators for independent evidence lanes.

Good candidates for parallelization:

• genetics versus expression evidence for the same gene or variant
• structure versus chemistry for the same target
• literature versus clinical evidence for the same disease
• multiple PheWAS cohorts for the same variant

Keep the coordinating agent responsible for:

• interpreting the user request and defining scope
• resolving identifiers and canonical entities
• reconciling conflicting evidence across sources
• writing the final synthesis

Each subagent should receive a bounded objective and return concise findings, caveats, sources used, and any artifact paths.

Evidence Quality Notes

• Cross-check orthogonal evidence types rather than over-indexing on one source.
• Call out ancestry limitations, tissue specificity, study design caveats, and evidence gaps explicitly.
• Treat heatmap breadth (Open Targets datasource scores) as evidence-source context, not proof of causality or direction of effect.
• For variant lookups, verify coordinate build (GRCh37 vs GRCh38) before querying cohort-specific PheWAS skills.