Latest information on COVID-19
Support teaching, research, and patient care.
Trained in the French school of Data Analysis in Montpellier, Susan Holmes has been working in non parametric multivariate statistics applied to Biology since 1985. She has taught at MIT, Harvard and was an Associate Professor of Biometry at Cornell before moving to Stanford in 1998. She created the Thinking Matters class: Breaking Codes and Finding patterns and likes working on big messy data sets, mostly from the areas of Immunology, Cancer Biology and Microbial Ecology. Her theoretical interests include applied probability, MCMC (Monte Carlo Markov chains), Graph Limit Theory, Differential Geometry and the topology of the space of Phylogenetic Trees. She wrote the book Modern Statistics for Modern Biology with Wolfgang Huber from EMBL and teaches the material as a crash course (BIOS221) regularly every year. Her current focus is improving the statistical analyses and reproducibility of data in perturbation studies of the Human Microbiome.
We are using statistical multivariate methods we are devloping new methods for modeling and visualizing the dynamics of bacterial community networks in the human microbiome.
Develop tools for testing evokutionary signals in bacterial data.
We use phylogenetic trees, networks and multivariate analyses to deompose the complexities of drug resistance in HIV.
Combining 16S sRNA data, metabolic and transcriptomic data to predict resilience in the human microbiome after perturbations.
Our work focuses on large heterogeneous multi-layer data analyses. Whether using image analysis and segmentation for the study of cancer and immune cell interactions, or brain imaging and DNA sequence analyses for the study of dependencies between genetic and neurological dynamics, all these statistical studies have involved large complex datasets of different types where dynamics of interactions between different components of a system are the key to understanding the underlying biology.We have generalized methods such as Principal Components Analysis (PCA) to more diverse data incorporating spatial information as well as tree dependency structures. This has proved useful in the study of drug resistant mutations in HIV and in the study of the dynamics of bacterial communities in the Human Microbiome.The statistical bases for these nonparametric methods are computer intensive methods using optimization and Kernels and we often find useful embeddings of high dimensional data in low dimensional structures, the extreme case being finding a natural ordering in high dimensional data. More general manifolds have also proved useful in one of our current projects, joint with Xavier Pennec of INRIA-SophiaAntiolis which focuses on the uses of differential geometry in computational anatomy and image processing.In a long term collaboration with Professor David Relman (Stanford Medical School) we are developing a multi-table toolbox of non parametric methods that enable users to normalize and visualize the multiple facets of the microbiota in the human body under different classes of perturbations. The tools developed in this project are all open source packages developed in R and provide an example of reproducible research in action.