Disentangled Feature Importance: A New Framework for Attributing Predictive Signals in Correlated Data
A research team has introduced Disentangled Feature Importance (DFI), a new statistical framework for attributing predictive signals across correlated input variables in machine learning models. Unlike existing conditional-incremental measures suited for feature selection, DFI uses entropic optimal transport to map covariates to an independent latent space before attributing importance back to original variables. The method addresses a gap in post-hoc model interpretation where correlated features share predictive information that standard approaches treat as redundancy rather than distributing it meaningfully.
The paper, posted to arXiv and spanning 29 main pages plus 44 supplementary pages, introduces Disentangled Feature Importance (DFI) as a population-level attribution framework designed specifically for settings where input predictors are statistically dependent. The authors argue that widely used conditional-incremental feature importance measures answer a fundamentally different question—targeting conditional incremental predictive value under squared-error loss—making them ill-suited for post-hoc interpretation when the goal is to distribute shared predictive signal across correlated measurement channels. DFI addresses this by mapping covariates to an independent latent representation using a specified entropic optimal transport geometry, computing importance in that latent space, and projecting attributions back to original covariates via barycentric sensitivities. The framework is shown to recover the classical R² decomposition for correlated regressors in the Gaussian linear case, providing a theoretically grounded connection to established statistics. The authors derive influence-function-based inference procedures under nuisance-rate and smoothness conditions, enabling uncertainty quantification for the attributions. Empirical validation includes simulations and an application to HIV-1 neutralization-resistance data, where DFI is reported to yield stable and interpretable attributions. The work spans machine learning, statistics theory, and methodology, and has undergone three revisions since its initial submission in June 2025.
What's missing
The paper is a preprint and has not yet undergone formal peer review, so its theoretical claims and empirical results have not been independently validated. Key open questions include computational scalability of the entropic optimal transport step to very high-dimensional settings, and whether the framework's performance advantages over alternatives hold across diverse real-world datasets beyond the HIV-1 application shown.
What different sources said
- arXiv cs.LGCenter
Disentangled Feature Importance
Related
Gut Bacteria Enzyme Found to Break Down Heat-Processed Food Compounds, Producing Novel Biogenic Amines
Researchers have discovered that an enzyme in common gut bacteria can degrade N-epsilon-carboxymethyllysine (CML), a compound formed during thermal food processing, producing previously unknown biogenic amines. The enzyme, ornithine decarboxylase SpeC from enterobacteria, acts on CML and related modified lysine derivatives through a low-level 'underground' catalytic activity. This finding suggests a previously unrecognized communication axis between thermally processed dietary compounds and gut microbial physiology, with potential implications for host health.
Full-Length Gene Sequencing Reveals Two Distinct Bacterial Communities in Black-Legged Ticks Expanding Into Canada
Researchers used Oxford Nanopore full-length 16S rRNA gene sequencing to characterize the microbiome of Ixodes scapularis black-legged ticks collected in Nova Scotia, Canada, distinguishing between tick-adapted bacteria and environmentally acquired bacteria. The study comes as I. scapularis — the primary vector of Lyme disease — is rapidly expanding northward into Canada due to climate change. The findings suggest that environmentally derived bacteria in tick microbiomes are not mere contamination, which has implications for how tick microbiome data is collected and interpreted across surveillance studies.
Study Identifies Metabolic Link Between Cell Envelope Stress and Biofilm Formation in Bacteria
Researchers have discovered that the metabolite acetyl-CoA directly inhibits enzymes that degrade the bacterial signaling molecule c-di-GMP, connecting cell envelope biosynthesis stress to biofilm formation in Pseudomonas aeruginosa. The study found that sub-inhibitory concentrations of antibiotics targeting early peptidoglycan biosynthesis — but not other antibiotic classes — elevate c-di-GMP levels by reducing phosphodiesterase activity, with acetyl-CoA competing for the enzyme active site. Because the relevant enzyme domain is broadly conserved across bacterial species, this checkpoint mechanism may be widespread and could have implications for understanding antibiotic-induced biofilm responses.