4th of April 2017
This Semovi is a joint event together with the CLARA and will be held during the FORUM DE LA RECHERCHE EN CANCÉROLOGIE AUVERGNE-RHÔNE-ALPES
You will have to register here
Registration to Semovi allows access to CLARA Forum conferences and social event.
Location: Espace Tête d'Or
|
14.15 - 15.15 |
David Gfeller University of Lausanne
|
Computational modeling of cancer-immune cells interactions
|
|---|---|---|
|
15.15 - 15.45 |
Coffee break |
|
|
15.45 - 16.15 |
Anthony Ferrari CRCL, |
A phenotypic and mechanistic perspective on heterogeneity of HER2-positive breast cancers |
|
16.15 - 16.45 |
Matthieu Foll CIRC
|
needlestack: an ultra-sensitive variant caller for multi-sample next generation sequencing data with applications to circulating tumor DNA detection. |
| 16.45 - 17.30 |
General discussion |
Computational modeling of cancer-immune cells interactions
Tumors are composed of various cell types, including cancer immune and stromal cells. Immune cells infiltrating the tumor microenvironment play a major role in shaping tumor progression, response to (immuno-)therapy and patient survival.In the first part of this seminar, I will discuss our ongoing work to use tumor gene expression data to better understand the immune contexture of tumors. In particular, I will introduce our novel algorithm to accurately Estimate the Proportion of Immune and Cancer cells (EPIC) from bulk tumor gene expression data and present some recent work on how single-cell RNA-Seq can unravel new properties of immune cell populations. In the second part, I will discuss our efforts to understand the molecular bases of cancer cell immune recognition. In particular, I will present a novel machine learning algorithm to train HLA-ligand predictors on in-depth HLA peptidomics data. Remarkably, our results show that these large and unbiased datasets of naturally presented peptides are ideal to train HLA-ligand predictors and result in very large improvement in neo-antigen predictions. Moreover, our unique dataset of HLA ligands enable us to unravel some of the determinants of HLA binding motifs, including an unexpected allosteric regulation of HLA specificity that we experimentally verify with mutagenesis and in vitro binding assays.
A phenotypic and mechanistic perspective on heterogeneity of HER2-positive breast cancers
Anthony Ferrari
Analysis of gene expression and whole-genome features of HER2 positive breast tumors supports the idea that their intrinsic heterogeneity actually reflects their cell of origin, suggesting that HER2 amplification is an embedded event in the natural history of these tumors. Possible mechanisms for this event involve breakage-fusion-bridge and chromothripsis.
needlestack: an ultra-sensitive variant caller for multi-sample next generation sequencing data with applications to circulating tumor DNA detection.
The comprehensive characterization of somatic mutations by screening cancer genomes can help to understand cancer appearance and progression but also to identify accurately predictive biomarkers. Circulating tumor DNA (ctDNA) is emerging as a key potential biomarker for the detection of pre- clinical cancer, through the assessment of target-gene mutations in the cell-free DNA (cfDNA). Nevertheless detecting somatic mutations in cancer genomes remains an unsolved problem, which is exacerbated when trying to identify mutations in ctDNA due to reduced variant allelic fractions. Indeed, Next Generation Sequencing (NGS) error level can reach this low proportion, and somatic variant calling from ctDNA is like finding a needle in a needlestack. Here we present the development and applications of needlestack, an ultra sensitive variant caller which estimates the distribution of sequencing errors directly from the data across multiple samples. Needlestack is based on the idea that analyzing large number of samples together can help estimate the distribution of sequencing errors to accurately identify variants present in very low proportion. At each position and for each candidate alteration, we model the sequencing error distribution using a robust negative binomial regression and detect variants as being outliers from this error model. Contrary to most existing algorithms, needlestack can deal with both single nucleotide substitutions (SNVs) and short insertions or deletions (INDELs), and is able to detect as low as 0.1% allelic fraction mutations.
You can find the poster here.
