prx — PROSPECT eXplore

Note

Part of the jx experiment - not an official PROSPECT project.

An experiment in agent-driven scientific data exploration, built around PROSPECT chemical-genetics data and Bond et al. 2025 — the reference-based MOA inference method that turns PROSPECT primary-screen data into mechanism-of-action assignments.

prx is a curated catalog of marimo notebooks for chemical-genetics analysis, plus a thin skill that lets an agent compose new analyses from them. Each notebook is both a runnable demonstration and a source of pure functions other notebooks can import and reuse directly. Given a new chemical-genetics question, the agent picks relevant notebooks, composes their functions into a new notebook, executes it in a live kernel, and hands back a self-contained, re-runnable result.

PROSPECT generates chemical-genetic interaction (CGI) profiles by screening compound libraries against pooled hypomorphic Mtb strains; Bond et al. 2025 introduced PCL (Perturbagen CLass) analysis — predict MOA for an unknown compound by comparing its CGI profile against a 437-compound annotated reference set.

The catalog

Each notebook ships with a committed session snapshot under notebooks/__marimo__/session/ so the molab preview renders cell outputs without re-executing.

Notebook	Role	Preview
nb01_orientation.py	Landing page, orientation, what's where
nb02_figshare_pull.py	Pull Bond et al. 2025 Figshare bundle, parse MOA and PCL annotations, build the reference-set spine
nb03_hypomorph_correlation.py	Load sGR GCT matrix, inspect strain-strain correlation across the 340-d CGI space
nb04_pretrained_baseline.py	Structure-only vs CGI-profile 1-NN baselines for MOA classification on the Bond reference set
nb05_collapse_diagnostic.py	Test whether same-MOA CGI similarity survives after controlling for pairwise chemical similarity
nb06_cgi_shape_diversity.py	PCL coverage, rarefaction, and effective CGI-shape diversity in the public Bond data

The machine-readable catalog table is catalog.toml's [[vignette]] blocks - each notebook, its reusable helpers, and what it does - which the vignette-catalog-compose-notebook skill reads.

Related public catalogs of the same pattern: jx for JUMP Cell Painting, fgx for FinnGenie human genetics, and dmx for DepMap Breadbox.

Getting started

The Bond et al. 2025 data is public - fetched from Figshare with pooch and SHA-256 pinned - so no API key or token is needed.

This catalog follows the vignette-catalog-skills pattern. The skill stores are gitignored, so a fresh clone has only skills-lock.json; restore the on-disk skill content first:

uv --version  # or: curl -LsSf https://astral.sh/uv/install.sh | sh
npx skills@1.5.20 add carpenter-singh-lab/vignette-catalog-skills -s vignette-catalog-compose-notebook -s vignette-catalog-scaffold -a claude-code -a codex -y
npx skills@1.5.20 add marimo-team/skills -s marimo-notebook -a claude-code -a codex -y
npx skills@1.5.20 add marimo-team/marimo-pair -s marimo-pair -a claude-code -a codex -y

Then open Claude Code or Codex in this repo and ask to get started. The vignette-catalog-compose-notebook skill launches nb01_orientation in a live marimo kernel and handles later analysis in the same workflow.

License

BSD 3-Clause — see LICENSE.