Are Tabular Foundation Models Robust to Realistic Query Distribution Shifts in Microbiome Data?

2026-06-23

AI Summary

This study systematically evaluates the robustness of tabular foundation models under distribution shifts in real-world microbiome data. To address the mismatch between support and query set distributions, the authors construct a benchmark comprising six gut microbiome datasets and introduce three biologically inspired perturbations: removal of high-abundance features, enhanced zero inflation, and injection of spurious non-zero entries. Experimental results demonstrate that all perturbations degrade model performance, with spurious non-zero injection causing the most severe decline. Notably, under zero-inflation perturbation, tabular foundation models perform significantly worse than a random forest baseline, revealing heightened sensitivity to disruptions in data sparsity structure. This work is the first to expose the vulnerability of such models in microbiome contexts, offering critical insights for developing more robust modeling approaches.

Abstract

Tabular foundation models (TFMs) achieve strong performance on microbiome abundance data, yet their robustness under realistic distribution shift remains poorly characterized. We introduce a benchmark that evaluates the robustness of TFMs to biologically inspired perturbations across six gut microbiome datasets spanning four disease contexts. In this in-context learning setting, models receive unperturbed support sets as context and are evaluated on perturbed query samples. To isolate robustness beyond "shortcut" features, we preserve the most discriminative taxa and apply three controlled perturbation strategies: (i) removal of high-abundance (uninformative) taxa, (ii) sparsification via increased zero-inflation, and (iii) zero-imputation via spurious non-zero injections. Our results show that protecting discriminative features is insufficient to guarantee stability under support-query shift: across datasets, all perturbations degrade model performance, with zero-imputation consistently the most harmful, indicating that corrupting global feature structure can break generalization even when key taxa are retained. Sparsification disproportionately affects TFMs relative to a classical random forest baseline, suggesting greater sensitivity to zero-inflation-type shifts. The code is publicly available at: https://github.com/UMMISCO/metagenomics-fm/.

Problem

• Tabular Foundation Models
• Distribution Shift
• Microbiome Data
• Robustness
• Zero-inflation

Innovation

• Tabular Foundation Models
• Distribution Shift
• Microbiome Data
• Zero-Inflation
• In-Context Learning