Information contracts determine what inverse perturbation benchmarks measure
Abstract
Inverse perturbation benchmarks rank interventions toward a desired cellular state, but high accuracy can answer a different question from biological prioritization. We present InverseBench, which makes candidate information and endpoint construction explicit, and FSTN, a linear pipeline evaluated under it. With identical inputs, candidates and validation, nested ridge reaches Spearman 0.959 for a transcriptional principal-component proxy but 0.445 for independently measured K562 growth fitness across 7,354 genes. Controlled response-entry availability across seven tasks gives heterogeneous gains, 0.442 for CD4 activation and 0.116 for external fitness. Yet predefined shortlist comparisons show no primary-cutoff penalty in K562 or a melanoma co-culture screen, and secondary cutoffs reverse the K562 ordering. FSTN's nonnegative set selector recovers 0.840 of measured two-gene constituents in three picks, against 0.771 for an unconstrained greedy. The framework separates proxy reconstruction, response-informed prioritization and unseen-intervention ranking: an endpoint-dependent accuracy gap does not by itself mean worse biological decisions.
Related
- Averaging in gene perturbation benchmarks hides perturbation-strength variation: Consequences for model evaluation
- CloneOT: predicting clone-specific perturbation responses with lineage-supervised neural inverse optimal transport
- Filling the Missing Puzzle: Integrating Interaction and Evolutionary Information to Enhance Protein Fitness Prediction in Inverse Folding
- A proof-carrying architecture for synthetic genetic logic circuits under stochastic temporal contracts
- Statistical patterns in the Ethereum blockchain: Analysis of EOAs and smart contracts in ERC20 token network
BibTeX
@misc{li2026information,
title = {{Information contracts determine what inverse perturbation benchmarks measure}},
author = {Jianfu Li and L. Frank Huang and Wenyu Song and Ying Li and James W. Jakub and Hong Qin and Yan Luo and Cui Tao},
howpublished = {bioRxiv},
year = {2026},
month = oct,
doi = {10.64898/2026.10.19.760271},
url = {https://www.biorxiv.org/content/10.64898/2026.10.19.760271v1},
note = {CC BY-NC},
}
RIS
TY - UNPB
TI - Information contracts determine what inverse perturbation benchmarks measure
AU - Jianfu Li
AU - L. Frank Huang
AU - Wenyu Song
AU - Ying Li
AU - James W. Jakub
AU - Hong Qin
AU - Yan Luo
AU - Cui Tao
PY - 2026
DA - 2026/10/19
PB - bioRxiv
DO - 10.64898/2026.10.19.760271
UR - https://www.biorxiv.org/content/10.64898/2026.10.19.760271v1
AB - Inverse perturbation benchmarks rank interventions toward a desired cellular state, but high accuracy can answer a different question from biological prioritization. We present InverseBench, which makes candidate information and endpoint construction explicit, and FSTN, a linear pipeline evaluated under it. With identical inputs, candidates and validation, nested ridge reaches Spearman 0.959 for a transcriptional principal-component proxy but 0.445 for independently measured K562 growth fitness across 7,354 genes. Controlled response-entry availability across seven tasks gives heterogeneous gains, 0.442 for CD4 activation and 0.116 for external fitness. Yet predefined shortlist comparisons show no primary-cutoff penalty in K562 or a melanoma co-culture screen, and secondary cutoffs reverse the K562 ordering. FSTN's nonnegative set selector recovers 0.840 of measured two-gene constituents in three picks, against 0.771 for an unconstrained greedy. The framework separates proxy reconstruction, response-informed prioritization and unseen-intervention ranking: an endpoint-dependent accuracy gap does not by itself mean worse biological decisions.
SN - 2692-8205
ER -