Learning Adaptive Perturbation-Conditioned Contexts for Robust Transcriptional Response Prediction

2026-02-21

AI Summary

Predicting high-dimensional transcriptional responses is challenging due to experimental noise and the sparsity of true gene effects, and existing methods often suffer from mean collapse, leading to high false-positive rates and poor interpretability. To address this, this work proposes AdaPert, a novel framework that learns perturbation-specific adaptive subgraphs from static biological knowledge graphs. By explicitly modeling structural priors and sparse signals within the graph, AdaPert effectively disentangles genuine transcriptional responses from noise. Integrating perturbation-conditioned modeling, adaptive subgraph learning, and knowledge graph incorporation, the method significantly outperforms current approaches across multiple genetic perturbation benchmarks. Notably, it achieves marked improvements in identifying differentially expressed genes (DEGs), offering both enhanced predictive accuracy and greater biological interpretability.

Technology Category

• Machine Learning: Graph-based Machine Learning
• Reasoning under Uncertainty: Graphical Models
• Data Mining & Knowledge Management: Graph Mining, Social Network Analysis & Community

Application Category

• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Predicting high-dimensional transcriptional responses to genetic perturbations is challenging due to severe experimental noise and sparse gene-level effects. Existing methods often suffer from mean collapse, where high correlation is achieved by predicting global average expression rather than perturbation-specific responses, leading to many false positives and limited biological interpretability. Recent approaches incorporate biological knowledge graphs into perturbation models, but these graphs are typically treated as dense and static, which can propagate noise and obscure true perturbation signals. We propose AdaPert, a perturbation-conditioned framework that addresses mean collapse by explicitly modeling sparsity and biological structure. AdaPert learns perturbation-specific subgraphs from biological knowledge graphs and applies adaptive learning to separate true signals from noise. Across multiple genetic perturbation benchmarks, AdaPert consistently outperforms existing baselines and achieves substantial improvements on DEG-aware evaluation metrics, indicating more accurate recovery of perturbation-specific transcriptional changes.

Problem

• transcriptional response prediction
• genetic perturbations
• mean collapse
• biological knowledge graphs
• sparsity

Innovation

• AdaPert
• perturbation-conditioned modeling
• sparsity-aware learning
• biological knowledge graph
• transcriptional response prediction