A latent factor approach for prediction from multiple assays

TL;DR

This paper introduces a novel latent factor model that captures both assay-specific and shared information across multiple measurement modalities, improving prediction accuracy in complex data settings.

Contribution

It proposes the Sparse Factor Method (SFM), an optimization-based approach that models latent factors for each assay and a common factor, addressing limitations of simple concatenation.

Findings

The SFM effectively captures assay-specific and shared structures.

The method improves prediction performance over traditional approaches.

An iterative algorithm efficiently solves the optimization problem.

Abstract

In many domains such as healthcare or finance, data often come in different assays or measurement modalities, with features in each assay having a common theme. Simply concatenating these assays together and performing prediction can be effective but ignores this structure. In this setting, we propose a model which contains latent factors specific to each assay, as well as a common latent factor across assays. We frame our model-fitting procedure, which we call the "Sparse Factor Method" (SFM), as an optimization problem and present an iterative algorithm to solve it.