Bayesian model-based method for clustering gene expression time series with multiple replicates

MMDPGP mixture modeling of transcriptomics time series with multiple replicates yields synthetic clusters with significantly higher purity than state-of-the-art model-based clustering approaches.

• Why it matters: Gene expression time series often include multiple replicates with different behaviors, and existing clustering methods may not model that replicate-specific structure well.
• What they did: The multiple models Dirichlet process Gaussian process (MMDPGP) mixture model samples Gaussian processes within an infinite mixture using a Dirichlet process, fitting separate GP models per replicate within each cluster.
• The result: In a case study on RNA-Seq time series of Clostridium botulinum with three experimental replicates, clusters from MMDPGP produced a larger number of significantly enriched gene sets than the other methods.

AI summary of the abstract · may contain errors · not medical advice · read the paper.