Cancer et génome: Bioinformatique, biostatistiques et épidémiologie des systèmes complexes

During the last 3 years, a number of approaches for the normalization of RNA sequencing data have emerged in the literature, differing both in the type of bias adjustment and in the statistical strategy adopted. However, as data continue to accumulate, there has been no clear consensus on the appropriate normalization method to be used or the impact of a chosen method on the downstream analysis. In this work, we focus on a comprehensive comparison of seven recently proposed normalization methods for the differential analysis of RNA-seq data, with an emphasis on the use of varied real and simulated datasets involving different species and experimental designs to represent data characteristics commonly observed in practice. Based on this comparison study, we propose practical recommendations on the appropriate normalization method to be used and its impact on the differential analysis of RNA-seq data.

We propose a convex-concave programming approach for the labeled weighted graph matching problem. The convex-concave programming formulation is obtained by rewriting the weighted graph matching problem as a least-square problem on the set of permutation matrices and relaxing it to two different optimization problems: a quadratic convex and a quadratic concave optimization problem on the set of doubly stochastic matrices. The concave relaxation has the same global minimum as the initial graph matching problem, but the search for its global minimum is also a hard combinatorial problem. We, therefore, construct an approximation of the concave problem solution by following a solution path of a convex-concave problem obtained by linear interpolation of the convex and concave formulations, starting from the convex relaxation. This method allows to easily integrate the information on graph label similarities into the optimization problem, and therefore, perform labeled weighted graph matching. The algorithm is compared with some of the best performing graph matching methods on four data sets: simulated graphs, QAPLib, retina vessel images, and handwritten Chinese characters. In all cases, the results are competitive with the state of the art.

In multi-task learning several related tasks are considered simultaneously, with the hope that by an appropriate sharing of information across tasks, each task may benefit from the others. In the context of learning linear functions for supervised classification or regression, this can be achieved by including a priori information about the weight vectors associated with the tasks, and how they are expected to be related to each other. In this paper, we assume that tasks are clustered into groups, which are unknown beforehand, and that tasks within a group have similar weight vectors. We design a new spectral norm that encodes this a priori assumption, without the prior knowledge of the partition of tasks into groups, resulting in a new convex optimization formulation for multi-task learning. We show in simulations on synthetic examples and on the IEDB MHC-I binding dataset, that our approach outperforms well-known convex methods for multi-task learning, as well as related non convex methods dedicated to the same problem.

INTRODUCTION: Basal-like carcinomas (BLCs) and human epidermal growth factor receptor 2 overexpressing (HER2+) carcinomas are the subgroups of breast cancers that have the most aggressive clinical behaviour. In contrast to HER2+ carcinomas, no targeted therapy is currently available for the treatment of patients with BLCs. In order to discover potential therapeutic targets, we aimed to discover deregulated signalling pathways in human BLCs. METHODS: In this study, we focused on the oncogenic phosphatidylinositol 3-kinase (PI3K) pathway in 13 BLCs, and compared it with a control series of 11 hormonal receptor negative- and grade III-matched HER2+ carcinomas. The two tumour populations were first characterised by immunohistochemistry and gene expression. The PI3K pathway was then investigated by gene copy-number analysis, gene expression profiling and at a proteomic level using reverse-phase protein array technology and tissue microarray. The effects of the PI3K inhibition pathway on proliferation and apoptosis was further analysed in three human basal-like cell lines. RESULTS: The PI3K pathway was found to be activated in BLCs and up-regulated compared with HER2+ tumours as shown by a significantly increased activation of the downstream targets Akt and mTOR (mammalian target of rapamycin). BLCs expressed significantly lower levels of the tumour suppressor PTEN and PTEN levels were significantly negatively correlated with Akt activity within that population. PTEN protein expression correlated significantly with PTEN DNA copy number and more importantly, reduced PTEN DNA copy numbers were observed specifically in BLCs. Similar to human samples, basal-like cell lines exhibited an activation of PI3K/Akt pathway and low/lack PTEN expression. Both PI3K and mTOR inhibitors led to basal-like cell growth arrest. However, apoptosis was specifically observed after PI3K inhibition. CONCLUSIONS: These data provide insight into the molecular pathogenesis of BLCs and implicate the PTEN-dependent activated Akt signalling pathway as a potential therapeutic target for the management of patients with poor prognosis BLCs.

We present a new clustering algorithm by proposing a convex relaxation of hierarchical clustering, which results in a family of objective functions with a natural geometric interpretation. We give efficient algorithms for calculating the continuous regularization path of solutions, and discuss relative advantages of the parameters. Our method experimentally gives state-of-the-art results similar to spectral clustering for non-convex clusters, and has the added benefit of learning a tree structure from the data. Contents 1

Regulation of RNA abundance and localization is a key step in gene expression control. Single-molecule RNA fluorescence in situ hybridization (smFISH) is a widely used single-cell-single-molecule imaging technique enabling quantitative studies of gene expression and its regulatory mechanisms. Today, these methods are applicable at a large scale, which in turn come with a need for adequate tools for data analysis and exploration. Here, we present FISH-quant v2, a highly modular tool accessible for both experts and non-experts. Our user-friendly package allows the user to segment nuclei and cells, detect isolated RNAs, decompose dense RNA clusters, quantify RNA localization patterns and visualize these results both at the single-cell level and variations within the cell population. This tool was validated and applied on large-scale smFISH image data sets, revealing diverse subcellular RNA localization patterns and a surprisingly high degree of cell-to-cell heterogeneity.

In multi-task learning several related tasks are considered simultaneously, with the hope that by an appropriate sharing of information across tasks, each task may benefit from the others. In the context of learning linear functions for supervised classification or regression, this can be achieved by including a priori information about the weight vectors associated with the tasks, and how they are expected to be related to each other. In this paper, we assume that tasks are clustered into groups, which are unknown beforehand, and that tasks within a group have similar weight vectors. We design a new spectral norm that encodes this a priori assumption, without the prior knowledge of the partition of tasks into groups, resulting in a new convex optimization formulation for multi-task learning. We show in simulations on synthetic examples and on the IEDB MHC-I binding dataset, that our approach outperforms well-known convex methods for multi-task learning, as well as related non-convex methods dedicated to the same problem. 1

BACKGROUND: Methylation of high-density CpG regions known as CpG Islands (CGIs) has been widely described as a mechanism associated with gene expression regulation. Aberrant promoter methylation is considered a hallmark of cancer involved in silencing of tumor suppressor genes and activation of oncogenes. However, recent studies have also challenged the simple model of gene expression control by promoter methylation in cancer, and the precise mechanism of and role played by changes in DNA methylation in carcinogenesis remains elusive. RESULTS: Using a large dataset of 672 matched cancerous and healthy methylomes, gene expression, and copy number profiles accross 3 types of tissues from The Cancer Genome Atlas (TCGA), we perform a detailed meta-analysis to clarify the interplay between promoter methylation and gene expression in normal and cancer samples. On the one hand, we recover the existence of a CpG island methylator phenotype (CIMP) with prognostic value in a subset of breast, colon and lung cancer samples, where a common subset of promoter CGIs hypomethylated in normal samples become hypermethylated. However, this hypermethylation is not accompanied by a decrease in expression of the corresponding genes, which are already lowly expressed in the normal genes. On the other hand, we identify tissue-specific sets of genes, different between normal and cancer samples, whose inter-individual variation in expression is significantly correlated with the variation in methylation of the 3' flanking regions of the promoter CGIs. These subsets of genes are not the same in the different tissues, nor between normal and cancerous samples, but transcription factors are over-represented in all subsets. CONCLUSION: Our results suggest that epigenetic reprogramming in cancer does not contribute to cancer development via direct inhibition of gene expression through promoter hypermethylation. It may instead modify how the expression of a few specific genes, particularly transcription factors, are associated with DNA methylation variations in a tissue-dependent manner.

Iron catalyses the oxidation of lipids in biological membranes and promotes a form of cell death called ferroptosis1. Defining where this chemistry occurs in the cell can inform the design of drugs capable of inducing or inhibiting ferroptosis in various disease-relevant settings. Genetic approaches have revealed suppressors of ferroptosis2–4; by contrast, small molecules can provide spatiotemporal control of the chemistry at work5. Here we show that the ferroptosis inhibitor liproxstatin-1 exerts cytoprotective effects by inactivating iron in lysosomes. We also show that the ferroptosis inducer RSL3 initiates membrane lipid oxidation in lysosomes. We designed a small-molecule activator of lysosomal iron—fentomycin-1—to induce the oxidative degradation of phospholipids and ultimately ferroptosis. Fentomycin-1 is able to kill iron-rich CD44high primary sarcoma and pancreatic ductal adenocarcinoma cells, which can promote metastasis and fuel drug tolerance. In such cells, iron regulates cell adaptation6,7 while conferring vulnerability to ferroptosis8,9. Sarcoma cells exposed to sublethal doses of fentomycin-1 acquire a ferroptosis-resistant cell state characterized by the downregulation of mesenchymal markers and the activation of a membrane-damage response. This phospholipid degrader can eradicate drug-tolerant persister cancer cells in vitro and reduces intranodal tumour growth in a mouse model of breast cancer metastasis. Together, these results show that control of iron reactivity confers therapeutic benefits, establish lysosomal iron as a druggable target and highlight the value of targeting cell states10. Some cancer cells exhibit high loads of reactive iron in lysosomes, and this feature is exploited by using fentomycin-1, a newly developed small molecule, to induce ferroptosis.

We present the group fused Lasso for detection of multiple change-points shared by a set of co-occurring one-dimensional signals. Change-points are detected by approximating the original signals with a constraint on the multidimensional total variation, leading to piecewise-constant approximations. Fast algorithms are proposed to solve the resulting optimization problems, either exactly or approximately. Conditions are given for consistency of both algorithms as the number of signals increases, and empirical evidence is provided to support the results on simulated and array comparative genomic hybridization data.

Analysing models of biological networks typically relies on workflows in which different software tools with sensitive parameters are chained together, many times with additional manual steps. The accessibility and reproducibility of such workflows is challenging, as publications often overlook analysis details, and because some of these tools may be difficult to install, and/or have a steep learning curve. The CoLoMoTo Interactive Notebook provides a unified environment to edit, execute, share, and reproduce analyses of qualitative models of biological networks. This framework combines the power of different technologies to ensure repeatability and to reduce users' learning curve of these technologies. The framework is distributed as a Docker image with the tools ready to be run without any installation step besides Docker, and is available on Linux, macOS, and Microsoft Windows. The embedded computational workflows are edited with a Jupyter web interface, enabling the inclusion of textual annotations, along with the explicit code to execute, as well as the visualization of the results. The resulting notebook files can then be shared and re-executed in the same environment. To date, the CoLoMoTo Interactive Notebook provides access to the software tools GINsim, BioLQM, Pint, MaBoSS, and Cell Collective, for the modeling and analysis of Boolean and multi-valued networks. More tools will be included in the future. We developed a Python interface for each of these tools to offer a seamless integration in the Jupyter web interface and ease the chaining of complementary analyses.

The Polycomb group of proteins is required for the proper orchestration of gene expression due to its role in maintaining transcriptional silencing. It is composed of several chromatin modifying complexes, including Polycomb Repressive Complex 2 (PRC2), which deposits H3K27me2/3. Here, we report the identification of a cofactor of PRC2, EZHIP (EZH1/2 Inhibitory Protein), expressed predominantly in the gonads. EZHIP limits the enzymatic activity of PRC2 and lessens the interaction between the core complex and its accessory subunits, but does not interfere with PRC2 recruitment to chromatin. Deletion of Ezhip in mice leads to a global increase in H3K27me2/3 deposition both during spermatogenesis and at late stages of oocyte maturation. This does not affect the initial number of follicles but is associated with a reduction of follicles in aging. Our results suggest that mature oocytes Ezhip-/- might not be fully functional and indicate that fertility is strongly impaired in Ezhip-/- females. Altogether, our study uncovers EZHIP as a regulator of chromatin landscape in gametes.

The canonical Wnt/β-catenin pathway is activated in triple-negative breast cancer (TNBC). The activation of this pathway leads to the expression of specific target genes depending on the cell/tissue context. Here, we analyzed the transcriptome of two different TNBC cell lines to define a comprehensive list of Wnt target genes. The treatment of cells with Wnt3a for 6h up-regulated the expression (fold change > 1.3) of 59 genes in MDA-MB-468 cells and 241 genes in HCC38 cells. Thirty genes were common to both cell lines. Beta-catenin may also be a transcriptional repressor and we found that 18 and 166 genes were down-regulated in response to Wnt3a treatment for 6h in MDA-MB-468 and HCC38 cells, respectively, of which six were common to both cell lines. Only half of the activated and the repressed transcripts have been previously described as Wnt target genes. Therefore, our study reveals 137 novel genes that may be positively regulated by Wnt3a and 104 novel genes that may be negatively regulated by Wnt3a. These genes are involved in the Wnt pathway itself, and also in TGFβ, p53 and Hedgehog pathways. Thorough characterization of these novel potential Wnt target genes may reveal new regulators of the canonical Wnt pathway. The comparison of our list of Wnt target genes with those published in other cellular contexts confirms the notion that Wnt target genes are tissue-, cell line- and treatment-specific. Genes up-regulated in Wnt3a-stimulated cell lines were more strongly expressed in TNBC than in luminal A breast cancer samples. These genes were also overexpressed, but to a much lesser extent, in HER2+ and luminal B tumors. We identified 72 Wnt target genes higher expressed in TNBCs (17 with a fold change >1.3) which may reflect the chronic activation of the canonical Wnt pathway that occurs in TNBC tumors.

Esthesioneuroblastoma (ENB) is a rare cancer of the olfactory mucosa, with no established molecular stratification to date. We report similarities of ENB with tumors arising in the neural crest and perform integrative analysis of these tumors. We propose a molecular-based subtype classification of ENB as basal or neural, both of which have distinct pathological, transcriptomic, proteomic, and immune features. Among the basal subtype, we uncovered an IDH2 R172 mutant-enriched subgroup (∼35%) harboring a CpG island methylator phenotype reminiscent of IDH2 mutant gliomas. Compared with the basal ENB methylome, the neural ENB methylome shows genome-wide reprogramming with loss of DNA methylation at the enhancers of axonal guidance genes. Our study reveals insights into the molecular pathogenesis of ENB and provides classification information of potential therapeutic relevance.

RNA localization is a crucial process for cellular function and can be quantitatively studied by single molecule FISH (smFISH). Here, we present an integrated analysis framework to analyze sub-cellular RNA localization. Using simulated images, we design and validate a set of features describing different RNA localization patterns including polarized distribution, accumulation in cell extensions or foci, at the cell membrane or nuclear envelope. These features are largely invariant to RNA levels, work in multiple cell lines, and can measure localization strength in perturbation experiments. Most importantly, they allow classification by supervised and unsupervised learning at unprecedented accuracy. We successfully validate our approach on representative experimental data. This analysis reveals a surprisingly high degree of localization heterogeneity at the single cell level, indicating a dynamic and plastic nature of RNA localization.

Abstract Circulating tumor cells (CTCs) and circulating tumor DNA (ctDNA) are two cancer-derived blood biomarkers that inform on patient prognosis and treatment efficacy in breast cancer. We prospectively evaluated the clinical validity of quantifying both CTCs (CellSearch) and ctDNA (targeted next-generation sequencing). Their combined value as prognostic and early monitoring markers was assessed in 198 HER2-negative metastatic breast cancer patients. All patients were included in the prospective multicenter UCBG study COMET (NCT01745757) and treated by first-line chemotherapy with weekly paclitaxel and bevacizumab. Blood samples were obtained at baseline and before the second cycle of chemotherapy. At baseline, CTCs and ctDNA were respectively detected in 72 and 74% of patients and were moderately correlated (Kendall’s τ = 0.3). Only 26 (13%) patients had neither detectable ctDNA nor CTCs. Variants were most frequently observed in TP53 and PIK3CA genes . KMT2C / MLL3 variants detected in ctDNA were significantly associated with a lower CTC count, while the opposite trend was seen with GATA3 alterations. Both CTC and ctDNA levels at baseline and after four weeks of treatment were correlated with survival. For progression-free and overall survival, the best multivariate prognostic model included tumor subtype (triple negative vs other), grade (grade 3 vs other), ctDNA variant allele frequency (VAF) at baseline (per 10% increase), and CTC count at four weeks (≥5CTC/7.5 mL). Overall, this study demonstrates that CTCs and ctDNA have nonoverlapping detection profiles and complementary prognostic values in metastatic breast cancer patients. A comprehensive liquid-biopsy approach may involve simultaneous detection of ctDNA and CTCs.

Xist is indispensable for X chromosome inactivation. However, how Xist RNA directs chromosome-wide silencing and why some regions are more efficiently silenced than others remains unknown. Here, we explore the function of Xist by inducing ectopic Xist expression from multiple different X-linked and autosomal loci in mouse aneuploid and female diploid embryonic stem cells in which Xist-mediated silencing does not lead to lethal functional monosomy. We show that ectopic Xist expression faithfully recapitulates endogenous X chromosome inactivation from any location on the X chromosome, whereas long-range silencing of autosomal genes is less efficient. Long interspersed elements facilitate inactivation of genes located far away from the Xist transcription locus, and genes escaping X chromosome inactivation show enrichment of CTCF on X chromosomal but not autosomal loci. Our findings highlight important genomic and epigenetic features acquired during sex chromosome evolution to facilitate an efficient X chromosome inactivation process.Xist RNA is required for X chromosome inactivation but it is not well understood how Xist silences some regions more efficiently than others. Here, the authors induce ectopic Xist expression from multiple different X-linked and autosomal loci in cells to explore Xist function.

BACKGROUND: To distinguish new primary breast cancers from true recurrences, pangenomic analyses of DNA copy number alterations (CNAs) using single-nucleotide polymorphism arrays have proven useful. METHODS: The pangenomic profiles of 22 pairs of primary breast carcinoma (ductal or lobular) and ipsilateral breast cancers from the same patients were analyzed. Hierarchical clustering was performed using CNAs and DNA breakpoint information. A partial identity score developed using DNA breakpoint information was used to quantify partial identities between two tumors. The nature of ipsilateral breast cancers (true recurrence vs new primary tumor) as defined using the clustering methods and the partial identity score was compared with that based on clinical characteristics. Metastasis-free survival was compared among patients with primary tumors and true recurrences as defined using the partial identity score and by clinical characteristics. All statistical tests were two-sided. RESULTS: All methods agreed on the nature of ipsilateral breast cancers for 14 pairs of samples. For five pairs, the clinical definition disagreed with both clustering methods. For three pairs, the two clustering methods were discordant and the one using DNA breakpoints agreed with the clinical definition. The partial identity score confirmed the nature of ipsilateral breast cancers as defined by clustering of DNA breakpoints in 21 of 22 pairs. The difference in metastasis-free survival of patients with new primary tumors and those with true recurrences was not statistically significant when tumors were defined based on clinical and histologic characteristics (5-year metastasis-free survival: 76%, 95% confidence interval [CI] = 52% to 100% for new primary tumors and 38%, 95% CI = 17% to 83% for true recurrences; P = .18; new primary tumor vs true recurrence, hazard ratio = 2.8, 95% CI = 0.6 to 13.7), but the difference was statistically significant when tumors were defined using the partial identity score (5-year metastasis-free survival: 100% for new primary tumors and 29%, 95% CI = 11% to 78% for true recurrences; P = .01). CONCLUSIONS: DNA breakpoint information more often agreed with the clinical determination than CNAs in this population. The partial identity score, which was calculated based on DNA breakpoints, allows statistical discrimination between new primary tumors and true recurrences that could outperform the clinical determination in terms of prognosis.

BACKGROUND: Nivolumab and pembrolizumab targeting programmed cell death protein 1 (PD-1) have recently been approved among patients with recurrent and/or metastatic head and neck squamous cell carcinoma (HNSCC) who failed platinum therapy. We aimed to evaluate the prognostic value of selected immune gene expression in HNSCC. PATIENTS AND METHODS: We retrospectively assessed the expression of 46 immune-related genes and immune-cell subpopulation genes including immune checkpoints by real-time polymerase chain reaction among 96 patients with HNSCC who underwent primary surgery at Institut Curie between 1990 and 2006. Univariate and multivariate analyses were performed to assess the prognostic value of dysregulated genes. RESULTS: The Median age of the population was 56 years [range: 35-78]. Primary tumour location was oral cavity (45%), oropharynx (21%), larynx (18%) and hypopharynx (17%). Twelve patients (13%) had an oropharyngeal human papillomavirus-positive tumour. Most significantly overexpressed immune-related genes were TNFRSF9/4-1BB (77%), IDO1 (75%), TNFSF4/OX40L (74%) and TNFRSF18/GITR (74%), and immune-cell subpopulation gene was FOXP3 (62%). Eighty-five percent of tumours analysed overexpressed actionable immunity genes, including PD-1/PD-L1, TIGIT, OX40/OX40L and/or CTLA4. Among the immune-related genes, high OX40L mRNA level (p = 0.0009) and low PD-1 mRNA level (p = 0.004) were associated with the highest risk of recurrence. Among the immune-cell subpopulation genes, patients with high PDGFRB mRNA level (p < 0.0001) and low CD3E (p = 0.0009) or CD8A mRNA levels (p = 0.004) were also at the highest risk of recurrence. CONCLUSIONS: OX40L and PDGFRB overexpression was associated with poor outcomes, whereas PD-1 overexpression was associated with good prognosis in patients with HNSCC treated with primary surgery, suggesting their relevance as potential prognostic biomarkers and major therapeutic targets.

We prove that the optimal assignment kernel, proposed recently as an attempt to embed labeled graphs and more generally tuples of basic data to a Hilbert space, is in fact not always positive definite.