Validating Interpretability in siRNA Efficacy Prediction: A Perturbation-Based, Dataset-Aware Protocol

2026-02-09

AI Summary

This study addresses the reliability of saliency maps in guiding siRNA sequence design, noting that their attribution quality lacks systematic validation and may lead to misleading edits. To tackle this, we propose a perturbation-based pre-synthesis validation protocol that, through counterfactual sensitivity testing and cross-dataset transfer analysis, identifies and defines two novel failure modes: “faithful yet incorrect” and “inverted saliency.” We advocate for saliency validation as an essential prerequisite for explanation-guided therapeutic design and introduce BioPrior, a biologically informed regularization method. Experiments across four benchmark datasets show that 19 out of 20 cross-validation runs pass our saliency validity test; BioPrior substantially improves attribution fidelity with only a minor, data-dependent trade-off in predictive performance.

Technology Category

• Search and Optimization: Metareasoning and Metaheuristics
• Machine Learning: Calibration & Uncertainty Quantification
• Constraint Satisfaction and Optimization: Satisfiability

Application Category

• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems
• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets

Abstract

Saliency maps are increasingly used as \emph{design guidance} in siRNA efficacy prediction, yet attribution methods are rarely validated before motivating sequence edits. We introduce a \textbf{pre-synthesis gate}: a protocol for \emph{counterfactual sensitivity faithfulness} that tests whether mutating high-saliency positions changes model output more than composition-matched controls. Cross-dataset transfer reveals two failure modes that would otherwise go undetected: \emph{faithful-but-wrong} (saliency valid, predictions fail) and \emph{inverted saliency} (top-saliency edits less impactful than random). Strikingly, models trained on mRNA-level assays collapse on a luciferase reporter dataset, demonstrating that protocol shifts can silently invalidate deployment. Across four benchmarks, 19/20 fold instances pass; the single failure shows inverted saliency. A biology-informed regularizer (BioPrior) strengthens saliency faithfulness with modest, dataset-dependent predictive trade-offs. Our results establish saliency validation as essential pre-deployment practice for explanation-guided therapeutic design. Code is available at https://github.com/shadi97kh/BioPrior.

Problem

• siRNA efficacy prediction
• interpretability validation
• saliency maps
• counterfactual sensitivity
• therapeutic design

Innovation

• saliency validation
• counterfactual sensitivity
• siRNA efficacy prediction
• BioPrior
• interpretability