System note · biomedical

Circuit Compass

Circuit Compass organizes competing explanations for a biological result and the evidence that could distinguish them. I built the analysis and handoff pipeline; the proposed mechanisms still need experimental validation.

System note · hypothesis-generating system

Artifacts: Implementation not currently public · access checked 10 Sep 2026
Reading time: 2 min read
My contribution: Designed and built the end-to-end analysis and handoff pipeline: data ingestion, hypothesis panels, pseudobulk analysis, literature/reagent layers, figures, and scientific self-critique.
Evidence: artifact-backed; biological interpretation provisional
Last checked: 13 Aug 2026
Provenance: This note describes the pipeline and its critique; the implementation is not currently public. Biological interpretations require independent and physical validation.

Decision-support loop

Do not collapse the paradox.
Make it testable.

ObservationBehavior ≠ molecular readout

Keep the contradiction alive.

Compensation

Desensitization

Activity elevation

Network failure

Public evidenceDataset + literature

Score directions; preserve cohort and modality limits.

Adversarial passCritique the story

Confounds · replication · statistics · alternative mechanisms

Reality-facing outputNext experiment

The handoff narrows what the wet lab should discriminate.

The system accelerates hypothesis generation. Authority remains with the experiment that can disagree.

Start with competing explanations

The project began with a collaborator’s experiment where behavior and molecular readouts appeared to disagree. The pipeline organizes possible explanations rather than selecting one plausible story immediately.

My part was computational. I wanted the handoff to make the alternatives inspectable: what supports each explanation, what else could produce the same observation, and which measurement would distinguish them. Choosing a plausible mechanism too early would make the software look more decisive than the evidence allowed.

What I built

I built the data-ingestion, analysis, literature, figure, and handoff workflow. It uses public single-nucleus datasets and aggregates measurements at the mouse level when biological replication permits that analysis.

The output brings hypotheses, supporting evidence, limitations, and proposed follow-up measurements into one reviewable document. Synthetic installation checks are kept separate from real-data outputs.

Where interpretation remains uncertain

The existing critique identifies cohort and sex/strain confounding, limits of replication and cell annotation, and the distance between RNA measurements and protein or circuit behavior. Several mechanisms may fit the same signature.

Those limitations determine what needs checking next. The pipeline does not establish a biological mechanism or substitute for the collaborator’s experiment.

What is available here

The implementation is not currently public. This page explains the workflow and its limits; it is not a reproducible release of the analysis.

This is a practical reason I care about independent verification. A computational workflow can organize more explanations than a team could easily work through by hand, but the usefulness depends on what happens at the handoff. I want a collaborator to be able to challenge the explanation and choose a measurement that could prove it wrong.

Source material · availability checked 11 September 2026

What can be inspected publicly

The implementation is not currently public. This page describes the workflow; it does not release collaborator data or establish a biological mechanism.