Domain-aware priors stabilize, not merely enable, vertical federated learning in data-scarce coral multi-omics

TL;DR

This paper demonstrates that domain-aware priors significantly enhance the stability and interpretability of vertical federated learning in data-scarce multi-omics coral stress classification, outperforming standard methods.

Contribution

It introduces REEF, a domain-aware VFL framework utilizing biologically motivated priors, which improves stability and performance in small-sample, high-dimensional biological data.

Findings

REEF achieves higher AUROC than baselines in coral stress classification.

Biological priors reduce variance and improve stability of VFL models.

Domain-informed dimensionality reduction is crucial for data-scarce regimes.

Abstract

Vertical federated learning (VFL) enables multi-laboratory collaboration on distributed multi-omics datasets without sharing raw data, but exhibits severe instability under extreme data scarcity (P >> N) when applied generically. Here, we investigate how domain-aware design choices; specifically gradient saliency-guided feature selection with biologically motivated priors; affect the stability, interpretability, and failure modes of VFL architectures in small-sample coral stress classification (N = 13 samples, P = 90,579 features across transcriptomics, proteomics, metabolomics, and microbiome data). We benchmark REEF (Robust Expert Encoder Federation), a domain-aware VFL framework, against two baselines on the Montipora capitata thermal stress dataset: (i) a standard NVFlare-based VFL and (ii) LASER, a state-of-the-art label-aware VFL method. REEF achieves an AUROC of 0.776 +/- 0.039 after reducing dimensionality by 98.6% (90,579 to 1,300 features), substantially outperforming NVFlare VFL at chance level (AUROC 0.500 +/- 0.125, p = 0.0106, Cohen's d = 2.265) and numerically exceeding LASER (AUROC 0.557 +/- 0.191, p = 0.0995, Cohen's d = 1.068), with 3-5-fold variance reduction. An equal-weights ablation confirms that biological priors specifically contribute stability: removing priors yields statistically indistinguishable mean AUROC (p = 0.405) but 2.3x higher variance (CV 0.110 vs 0.050). Negative control experiments using permuted labels produce AUROC near or below chance (0.357 for REEF, 0.238 for NVFlare), consistent with the absence of gross data leakage. These results motivate design principles for VFL in extreme P >> N regimes, emphasizing domain-informed dimensionality reduction, stability-focused evaluation, and interpretable feature selection for scarce biological data.