Biosystems Modeling and Machine Learning
Making sense of microbiome data requires complex machine learning and computational approaches that account for the non-normal nature of sequencing data, sparsity and irregularity of temporal sampling, and for having number of features often much larger than available samples. Additionally, these approaches need to consider in their predictions the underlying network structure (phylogenetic, metabolic pathways) that characterizes multispecies microbial ecosystems. PMD members have developed several computational techniques to infer how groups of microbes (or metabolic signatures) predict clinical outcome. These tools are applied to different large scale clinical studies and have allowed to identify novel biology on how the microbiome interacts with the immune system and the external environment.
Top Publications:
- Social interaction, noise and antibiotic-mediated switches in the intestinal microbiota.
- Ecological modeling from time-series inference: insight into dynamics and stability of intestinal microbiota
- MDSINE: Microbial Dynamical Systems INference Engine for microbiome time-series analyses
- Phage mobility is a core determinant of phage-bacteria coexistence in biofilms