Pachter lab reanalyzes MoTrPAC rat exercise data, finding label swaps, isoform signals, and a virus
lpachter · x · 2026-09-08
Lior Pachter and Conrad Oakes published a reanalysis of the MoTrPAC endurance-exercise rat dataset in BMC Genomic Data, highlighting why rigorous bioinformatics and statistics matter:
- Label swaps: the original data contained overlooked sample label swaps.
- Prediction: gene expression can accurately predict training volume; a GLM worked while an scVI latent space did not help.
- Isoform-level findings: ignoring isoforms is routine (MoTrPAC and various AI foundation models do it too), but drilling down reveals signals masked at gene level — e.g., a SUMO transcript clearly differential with exercise while the gene is not.
- A hidden virus: viral detection identified a sick rat, showing health is a key covariate — Tigit appeared exercise-differential but was actually an immune response to sickness.
The paper is open access, with full results reproducible via Google Colab notebooks.
More from Research
- OpenAI Foundation funds UNC tumor antigen project, hiring senior experimental scientist — iskander · 2026-10-27
- Steering attention heads toward honesty: up moral patienthood, down psychopathy — edelwax · 2026-10-27
- Are transcoders overkill? New paper argues plain neurons suffice for circuit analysis — burny_tech · 2026-10-27
- EMNLP paper GENIE: a fine-grained metric locating where LLM stories are novel — gregd_nlp · 2026-10-27
- Hashed n-gram value embeddings push GPT to 3.2115 loss, beating 30M track record — RobertTLange · 2026-10-27
- MissionReport benchmark: frontier models score just ~60% on government status reporting — Exp_Mark · 2026-10-27