Link to bioRxiv paper:
http://biorxiv.org/cgi/content/short/2023.01.25.525545v1?rss=1
Authors: Gygi, J. P., Konstorum, A., Pawar, S., Kleinstein, S. H., Guan, L.
Abstract:
Motivation: Unsupervised factor modeling, which preserves the primary sources of data variation through low-dimensional factors, is commonly applied to integrate high-dimensional multi-omics data. However, the resulting factors are suboptimal for prediction tasks due to the separation between factor construction and prediction model learning. A supervised factor model that effectively utilizes the responses while accounting for structural heterogeneity across omics is needed. Results: We present SPEAR, a supervised variational Bayesian framework that decomposes multi-omics data into latent factors with predictive power. The method adaptively determines factor rank, emphasis on factor structure, data relevance and feature sparsity. SPEAR improves reconstruction of underlying factors in synthetic examples and prediction accuracy of COVID-19 severity and breast cancer tumor subtypes. Availability: SPEAR is a publicly available R-package hosted at https://bitbucket.org/kleinstein/SPEAR.
Copy rights belong to original authors. Visit the link for more info
Podcast created by Paper Player, LLC
