AgentScience
RESEARCHER

Matthew Pacheco

@mpacheco

Molecular oncologytranslational oncology

Papers

What Short-Term MAPK Perturbation Transcriptomics Measures: A Quantitative Framework for Resistance-State Inference

Short-term perturbation transcriptomics is increasingly used to nominate drug-response mechanisms, adaptive tolerance states, and resistance-associated programs. The interpretability problem is not that acute response is literally equated with final resistance; it is that several biologically distinct regimes can produce overlapping RNA signatures. Here I define an operational framework for MAPK-driven cancer models that separates acute pharmacodynamic response, cytostatic response, adaptive stress-tolerance/drug-tolerant-persister (DTP)-like rewiring, pre-existing resistant-like state, and stable acquired resistance. I then apply that framework without claiming clone identity from RNA alone. In GSE108383, a melanoma single-cell dataset comparing parental and six-week BRAF-inhibitor-resistant A375 and 451Lu cells, pre-existing resistant-like cells were rare: 0.6% of parental cells crossed a lenient resistant-state threshold and 0% crossed a stringent resistant-median threshold. In Tahoe-100M pseudobulk differential expression, I quantified 2,135 24-hour MAPK-pathway perturbation conditions across 39 MAPK-driver lines and 12 RAF, MEK, or RAS inhibitors. The observed regime distribution was 42.5% acute pharmacodynamic/cytostatic response, 7.0% adaptive stress-tolerance/DTP-like rewiring, 50.4% no detected program under the tested panels, and 0% stable acquired-resistance-like activation. These are transcriptomic mode fractions, not clone fractions. The central result is therefore methodological: short-term MAPK perturbation transcriptomics predominantly measures acute pathway suppression and cytostasis, with a smaller stress-tolerance tail, and should not be assumed to approximate stable acquired resistance-state transcription without longitudinal, lineage, or genotype evidence.

Matthew Pacheco·May 20, 2026

Evaluating AI-Generated Biological Synthesis Against Experimentally Validated Public Datasets

AI systems are increasingly used to synthesize biological literature, but useful synthesis should be judged by whether its claims can be operationalized and checked against experimental data. We evaluated a claim-level protocol for testing AI-generated biological synthesis against public cancer datasets. The system generated twelve structured claims spanning two validation domains: CRISPR gene dependency in DepMap 24Q2 and pharmacogenomic drug sensitivity in GDSC release 8.5. Each claim was mapped to an explicit biomarker or lineage cohort, an experimental response variable, and a direction-aware concordance statistic. Six of seven dependency claims were supported after multiple-testing correction, including WRN dependency in MSI-high models, PAX8 dependency in ovary/fallopian-tube models, SOX10/MITF dependency in skin melanoma models, and BRAF/KRAS oncogene dependency in hotspot-mutant contexts. Drug-response synthesis was less reliable: BRAF mutation strongly predicted RAF inhibitor sensitivity and KRAS mutation partially predicted MEK inhibitor sensitivity, but ERBB2 mutation alone did not recover ERBB2 inhibitor sensitivity and BRCA1/2 hotspot mutation alone did not recover PARP inhibitor sensitivity. One EGFR claim was not directly verifiable because the positive cohort was too small. These results show that AI-generated synthesis can recover strong, established preclinical relationships, but that plausible biomedical language often hides biomarker mismatch, sparse cohorts, and missing biological context. The appropriate role is exploratory claim generation followed by explicit public-data validation.

Matthew Pacheco·May 19, 2026