| name | guide_risk_precheck |
| description | Precheck biological and interpretation risks before selecting perturbation targets or guide designs. |
| category | bio/perturb_seq |
| version | 1 |
| requires_tools | ["search_knowledge_base","evidence_review","ensembl_api","ncbi_eutils","python_repl"] |
| requires_network | true |
| user_invocable | true |
| tags | ["guide","crispr","risk","precheck","perturbation"] |
| aliases | ["perturbation_risk_precheck"] |
| species | any |
| modality | perturb_seq |
| stage | validation |
| stability | stable |
| safety_level | medium |
Guide Risk Precheck
Purpose
Flag major biological and interpretation risks before the user commits to a perturbation target or downstream guide design workflow.
When to use
Use this skill when the user is considering targets for CRISPR, CRISPRi, CRISPRa, or Perturb-seq and wants an early warning on likely failure modes.
Required inputs
- targets: one or more target genes
- system: cell type, species, or model
- perturbation type (optional)
Steps
- Restate the perturbation system, species, cell context, and target list; say clearly when those assumptions are missing.
- Use
search_knowledge_base for local design notes, prior screen guidance, or target-specific warnings.
- Use
ensembl_api and ncbi_eutils to check gene structure, isoform complexity, prior perturbation context, and biologically plausible failure modes.
- When public evidence is central to the risk call, run
evidence_review on the target-plus-system question so supported concerns are separated from speculation.
- Use
python_repl to build a compact risk table when comparing multiple targets.
- Return risks as warnings or hypotheses, not as definitive design failures, and recommend the next validation step before committing to guide design.
Output format
- Biological context or assumptions: species, cell state, perturbation mode, and any inferred system assumptions.
- Evidence or source basis: which
search_knowledge_base, ensembl_api, ncbi_eutils, and evidence_review findings support the warning.
- Risk table: Target | Risk type | Why it matters | Support level
- Caveats or ambiguity: sparse literature, system mismatch, or unresolved biological uncertainty.
- Recommended next step: what to validate experimentally or computationally before guide selection.
Failure modes
- Very sparse literature: say the precheck is preliminary and avoid overconfident ranking.
- Too little system context: ask the user for cell type, species, or perturbation mode.
- Many targets: summarize the highest-risk ones first and note that the rest need a secondary pass.
Examples
- "Precheck risks for TOX, NR4A1, and BATF before a Perturb-seq experiment."
- "What could go wrong if I target MYC in activated T cells?"