critique-hypothesis

name	critique_hypothesis
description	Critically evaluate a perturbation hypothesis — challenge assumptions, propose negative controls, and flag confounders.
category	bio/perturb_seq
version	1
requires_tools	["search_knowledge_base","ncbi_eutils","fetch_url","python_repl"]
requires_network	true
user_invocable	true
species	any
modality	perturb_seq
stage	validation
stability	evolving
safety_level	medium

Critique Hypothesis

Purpose

Challenge an existing perturbation hypothesis (from generate_perturbation_hypothesis or user-provided) to improve experimental rigor.

When to use

User says "critique this hypothesis", "what could go wrong with this experiment?", or "play devil's advocate."

Inputs

• hypothesis: Text of the hypothesis (pasted or from previous skill output).
• target_gene (optional): Gene being perturbed.
• cell_type (optional): Cell type used.

---

Steps

Step 1 — Identify key assumptions

Read the hypothesis and list the major assumptions:

1. That the gene is the primary driver of the phenotype.
2. That the perturbation method achieves complete loss/gain of function.
3. That the phenotype is specific to this perturbation (no broad toxicity).
4. That the cell type/model is appropriate.

Step 2 — Literature check for contradictions

Use ncbi_eutils (esearch db=pubmed) to search for:

• "{gene} {cell_type} off-target effect"
• "{gene} redundant paralog"
• "{gene} essential gene toxicity"

Retrieve titles; flag any papers that contradict or complicate the hypothesis.

Step 3 — Local knowledge check

Use search_knowledge_base for lab notes, past failed experiments, or known cell-line quirks.

Step 4 — Structured critique

For each assumption, write a brief challenge:

## Challenge: {Assumption}
- **What could go wrong**: Specific risk (off-target, redundancy, toxicity, indirect effect).
- **Evidence**: PMID or local source, or "not found in literature."
- **Mitigation**: How to address this (e.g. rescue experiment, paralog knockout, dose titration).

Step 5 — Negative controls

Propose negative controls:

• Non-targeting sgRNA (essential baseline).
• Perturbation of a gene in the same pathway but with an opposite expected phenotype.
• A second guide RNA targeting the same gene (to rule out on-target guide effects).

Step 6 — Confounders to watch

• Cell cycle effects (perturbing cell cycle regulators creates indirect transcriptional changes).
• Batch effects if multiple guides are pooled.
• MOI (multiplicity of infection) if viral delivery.
• Guide efficiency variation (measure KO efficiency by Western or sequencing).

---

Output format

• Assumption list: Numbered.
• Per-assumption challenge: As in Step 4 format.
• Negative controls table: Control | Purpose.
• Confounders list: Bullet list.
• Verdict: "Strong hypothesis with caveats", "Moderate — needs more evidence", or "Weak — reconsider target."