The development of DNA sequencing technologies allows us to collect large amounts of molecular data about the genome of each individual, and opens the possibility to predict drug response or evaluate the risk of various diseases from one's molecular identity. It also raises statistical and computational challenges, as the quantity of data collected per sample is usually far larger than the number of samples available to estimate predictive models. In this talk I will discuss some of these challenges and describe a few methods to attack them through regularization or change of representation, illustrated on applications in cancer prognosis and drug response prediction from gene expression and somatic mutations

Sexual reproduction is an extraordinarily widespread phenomenon that assures the production of new genetic combinations in nearly all eukaryotic lineages. Although the core features of sexual reproduction (meiosis and syngamy) are highly conserved, the control mechanisms that determine whether an individual is male or female are remarkably labile across eukaryotes. In genetically controlled sexual systems, gender is determined by sex chromosomes, which have emerged independently and repeatedly during evolution. Sex chromosomes have been studied in only a handful of classical model organism, and empirical knowledge on the origin and evolution of the sexes is still surprisingly incomplete. I will present how our group is exploiting the richness of sexual characteristics of the brown algae to gain novel insights into the functional and evolutionary interactions between the sex chromosomes and reproductive/life cycle features.

RNA viruses rapidly evade the host immune responses by changing targeted proteins. I will discuss how deep sequencing of HIV populations from many consecutive samples can be used to track virus evolution and quantify competing tendencies of immune escape and rapid replication. In a similar manner, the rapid evolution of influenza viruses necessitates the frequent updates of the seasonal influenza vaccines. We have used models of rapidly evolving populations to predict which strains of influenza virus are most likely to dominate future influenza seasons. Such models can inform the selection of virus strains used to produce influenza vaccines.

Numerous studies have documented the rapid evolution of mammalian transcriptional regulation, especially enhancers, and yet gene expression patterns are often conserved between species. How stable gene expression is maintained by rapidly evolving regulatory landscapes is a fundamental question in evolutionary genetics.We tested the contributions of both complexity and conservation of regulatory elements on gene expression evolution, using an integrated dataset of active promoters, enhancers and gene expression output from the same liver samples across twenty mammalian species. Our methodology captures regulatory activities that range from essential to dispensable, and from highly-conserved across mammals to present in only one species. This analysis revealed that gene expression levels and stability are reflected by the complexity of their regulatory landscape, both within a single species and across mammals. Regulatory activities conserved across placental mammals exert a powerful stabilizing effect, associating with gene expression levels that are simultaneously high and evolutionarily stable. In contrast, recently-evolved enhancers…

We introduce an accurate and parameter free approach to estimate the probability density for non-uniform samples embedded in high dimensional and curved manifolds.The estimator uses only distances between the points and not their coordinates and does not require specifying the functional form of the probability. Moreover, it provides an approximately unbiased measure of the uncertainty of the estimate.The availability of an error estimate allows distinguishing genuine probability peaks from statistical fluctuations due to finite sampling. This is an essential ingredient for  finding the meaningful  peaks and the saddles between them.  Based on this information, we then reconstruct a hierarchical topography of the probability density, which provides a compact representation of the data set.We demonstrate the usefulness of the approach in classification problems in functional magnetic resonance imaging, proteomics and metagenomics.

Numerous studies have documented the rapid evolution of mammalian transcriptional regulation, especially enhancers, and yet gene expression patterns are often conserved between species. How stable gene expression is maintained by rapidly evolving regulatory landscapes is a fundamental question in evolutionary genetics.We tested the contributions of both complexity and conservation of regulatory elements on gene expression evolution, using an integrated dataset of active promoters, enhancers and gene expression output from the same liver samples across twenty mammalian species. Our methodology captures regulatory activities that range from essentialto dispensable, and from highly-conserved across mammals to present in only one species. This analysis revealed that gene expression levels and stability are reflected by the complexity of their regulatory landscape, both within a single species and across mammals. Regulatory activities conserved across placental mammals exert a powerful stabilizing effect, associating with gene expression levels that are simultaneously high and evolutionarily stable. In contrast,recently-evolved enhancers contribute weakly to gene expression and transcriptional stability, consistent with a model whereby a sizable fraction…

Cancer is a disease that is mostly driven by somatic mutations accumulating in the genome during an individual’s lifetime. Recent advances in DNA sequencing technology have enabled genome-wide measurements of these mutations in large cohorts of cancer patients. A major challenge in analyzing these data is to distinguish functional "driver" mutations responsible for cancer progression from “passenger”, random mutations not related to the disease. Recent cancer sequencing studies have shown that somatic mutations are distributed over a large number of genes. This mutational heterogeneity is due in part to the fact that somatic mutations target pathways, or groups of genes, and that a mutation in any of dozens possible genes might be sufficient to perturb a pathway. While some of the cancer driver pathways are well characterized, many others are only approximately known. I will describe algorithms for discovering cancer driver pathways using DNA mutation data from large cohorts of cancer samples.…