Feature selection by replicate reproducibility and non-redundancy.

MOTIVATION: A fundamental step in many analyses of high-dimensional data is dimension reduction. Two basic approaches are introduction of new synthetic coordinates and selection of extant features. Advantages of the latter include interpretability, simplicity, transferability, and modularity. A common criterion for unsupervized feature selection is variance or dynamic range. However, in practice, it can occur that high-variance features are noisy, that important features have low variance, or that variances are simply not comparable across features because they are measured in unrelated numeric scales or physical units. Moreover, users may want to include measures of signal-to-noise ratio and non-redundancy into feature selection. RESULTS: Here, we introduce the RNR algorithm, which selects features based on (i) the reproducibility of their signal across replicates and (ii) their non-redundancy, measured by linear dependence. It takes as input a typically large set of features measured on a collection of objects with two or more replicates per object. It returns an ordered list of features, i1,i2,…,ik, where feature i1 is the one with the highest reproducibility across replicates, i2 that with the highest reproducibility across replicates after projecting out the dimension spanned by i1, and so on. Applications to microscopy-based imaging of cells and proteomics highlight benefits of the approach. AVAILABILITY AND IMPLEMENTATION: The RNR method is available via Bioconductor (Huber W, Carey VJ, Gentleman R et al. (Orchestrating high-throughput genomic analysis with bioconductor. Nat Methods 2015;12:115-21.) in the R package FeatSeekR. Its source code is also available at https://github.com/tcapraz/FeatSeekR under the GPL-3 open source license.

A single-sample workflow for joint metabolomic and proteomic analysis of clinical specimens.

Understanding the interplay of the proteome and the metabolome helps to understand cellular regulation and response. To enable robust inferences from such multi-omics analyses, we introduced and evaluated a workflow for combined proteome and metabolome analysis starting from a single sample. Specifically, we integrated established and individually optimized protocols for metabolomic and proteomic profiling (EtOH/MTBE and autoSP3, respectively) into a unified workflow (termed MTBE-SP3), and took advantage of the fact that the protein residue of the metabolomic sample can be used as a direct input for proteome analysis. We particularly evaluated the performance of proteome analysis in MTBE-SP3, and demonstrated equivalence of proteome profiles irrespective of prior metabolite extraction. In addition, MTBE-SP3 combines the advantages of EtOH/MTBE and autoSP3 for semi-automated metabolite extraction and fully automated proteome sample preparation, respectively, thus advancing standardization and scalability for large-scale studies. We showed that MTBE-SP3 can be applied to various biological matrices (FFPE tissue, fresh-frozen tissue, plasma, serum and cells) to enable implementation in a variety of clinical settings. To demonstrate applicability, we applied MTBE-SP3 and autoSP3 to a lung adenocarcinoma cohort showing consistent proteomic alterations between tumour and non-tumour adjacent tissue independent of the method used. Integration with metabolomic data obtained from the same samples revealed mitochondrial dysfunction in tumour tissue through deregulation of OGDH, SDH family enzymes and PKM. In summary, MTBE-SP3 enables the facile and reliable parallel measurement of proteins and metabolites obtained from the same sample, benefiting from reduced sample variation and input amount. This workflow is particularly applicable for studies with limited sample availability and offers the potential to enhance the integration of metabolomic and proteomic datasets.

SpatialData: an open and universal data framework for spatial omics.

Spatially resolved omics technologies are transforming our understanding of biological tissues. However, the handling of uni- and multimodal spatial omics datasets remains a challenge owing to large data volumes, heterogeneity of data types and the lack of flexible, spatially aware data structures. Here we introduce SpatialData, a framework that establishes a unified and extensible multiplatform file-format, lazy representation of larger-than-memory data, transformations and alignment to common coordinate systems. SpatialData facilitates spatial annotations and cross-modal aggregation and analysis, the utility of which is illustrated in the context of multiple vignettes, including integrative analysis on a multimodal Xenium and Visium breast cancer study.

In situ analysis of osmolyte mechanisms of proteome thermal stabilization.

Organisms use organic molecules called osmolytes to adapt to environmental conditions. In vitro studies indicate that osmolytes thermally stabilize proteins, but mechanisms are controversial, and systematic studies within the cellular milieu are lacking. We analyzed Escherichia coli and human protein thermal stabilization by osmolytes in situ and across the proteome. Using structural proteomics, we probed osmolyte effects on protein thermal stability, structure and aggregation, revealing common mechanisms but also osmolyte- and protein-specific effects. All tested osmolytes (trimethylamine N-oxide, betaine, glycerol, proline, trehalose and glucose) stabilized many proteins, predominantly via a preferential exclusion mechanism, and caused an upward shift in temperatures at which most proteins aggregated. Thermal profiling of the human proteome provided evidence for intrinsic disorder in situ but also identified potential structure in predicted disordered regions. Our analysis provides mechanistic insight into osmolyte function within a complex biological matrix and sheds light on the in situ prevalence of intrinsically disordered regions.

Multimodal and spatially resolved profiling identifies distinct patterns of T cell infiltration in nodal B cell lymphoma entities.

The redirection of T cells has emerged as an attractive therapeutic principle in B cell non-Hodgkin lymphoma (B-NHL). However, a detailed characterization of lymphoma-infiltrating T cells across B-NHL entities is missing. Here we present an in-depth T cell reference map of nodal B-NHL, based on cellular indexing of transcriptomes and epitopes, T cell receptor sequencing, flow cytometry and multiplexed immunofluorescence applied to 101 lymph nodes from patients with diffuse large B cell, mantle cell, follicular or marginal zone lymphoma, and from healthy controls. This multimodal resource revealed quantitative and spatial aberrations of the T cell microenvironment across and within B-NHL entities. Quantitative differences in PD1+ TCF7- cytotoxic T cells, T follicular helper cells or IKZF3+ regulatory T cells were linked to their clonal expansion. The abundance of PD1+ TCF7- cytotoxic T cells was associated with poor survival. Our study portrays lymphoma-infiltrating T cells with unprecedented comprehensiveness and provides a unique resource for the investigation of lymphoma biology and prognosis.

Improved discovery of RNA-binding protein binding sites in eCLIP data using DEWSeq.

Enhanced crosslinking and immunoprecipitation (eCLIP) sequencing is a method for transcriptome-wide detection of binding sites of RNA-binding proteins (RBPs). However, identified crosslink sites can deviate from experimentally established functional elements of even well-studied RBPs. Current peak-calling strategies result in low replication and high false positive rates. Here, we present the R/Bioconductor package DEWSeq that makes use of replicate information and size-matched input controls. We benchmarked DEWSeq on 107 RBPs for which both eCLIP data and RNA sequence motifs are available and were able to more than double the number of motif-containing binding regions relative to standard eCLIP processing. The improvement not only relates to the number of binding sites (3.1-fold with known motifs for RBFOX2), but also their subcellular localization (1.9-fold of mitochondrial genes for FASTKD2) and structural targets (2.2-fold increase of stem-loop regions for SLBP. On several orthogonal CLIP-seq datasets, DEWSeq recovers a larger number of motif-containing binding sites (3.3-fold). DEWSeq is a well-documented R/Bioconductor package, scalable to adequate numbers of replicates, and tends to substantially increase the proportion and total number of RBP binding sites containing biologically relevant features.

The changing career paths of PhDs and postdocs trained at EMBL.

Individuals with PhDs and postdoctoral experience in the life sciences can pursue a variety of career paths. Many PhD students and postdocs aspire to a permanent research position at a university or research institute, but competition for such positions has increased. Here, we report a time-resolved analysis of the career paths of 2284 researchers who completed a PhD or a postdoc at the European Molecular Biology Laboratory (EMBL) between 1997 and 2020. The most prevalent career outcome was Academia: Principal Investigator (636/2284=27.8% of alumni), followed by Academia: Other (16.8%), Science-related Non-research (15.3%), Industry Research (14.5%), Academia: Postdoc (10.7%) and Non-science-related (4%); we were unable to determine the career path of the remaining 10.9% of alumni. While positions in Academia (Principal Investigator, Postdoc and Other) remained the most common destination for more recent alumni, entry into Science-related Non-research, Industry Research and Non-science-related positions has increased over time, and entry into Academia: Principal Investigator positions has decreased. Our analysis also reveals information on a number of factors - including publication records - that correlate with the career paths followed by researchers.

MsQuality: an interoperable open-source package for the calculation of standardized quality metrics of mass spectrometry data.

MOTIVATION: Multiple factors can impact accuracy and reproducibility of mass spectrometry data. There is a need to integrate quality assessment and control into data analytic workflows. RESULTS: The MsQuality package calculates 43 low-level quality metrics based on the controlled mzQC vocabulary defined by the HUPO-PSI on a single mass spectrometry-based measurement of a sample. It helps to identify low-quality measurements and track data quality. Its use of community-standard quality metrics facilitates comparability of quality assessment and control (QA/QC) criteria across datasets. AVAILABILITY AND IMPLEMENTATION: The R package MsQuality is available through Bioconductor at https://bioconductor.org/packages/MsQuality.

Ex vivo drug response profiling for response and outcome prediction in hematologic malignancies: the prospective non-interventional SMARTrial.

Ex vivo drug response profiling is a powerful tool to study genotype-drug response associations and is being explored as a tool set for precision medicine in cancer. Here we conducted a prospective non-interventional trial to investigate feasibility of ex vivo drug response profiling for treatment guidance in hematologic malignancies (SMARTrial, NCT03488641 ). The primary endpoint to provide drug response profiling reports within 7 d was met in 91% of all study participants (N = 80). Secondary endpoint analysis revealed that ex vivo resistance to chemotherapeutic drugs predicted chemotherapy treatment failure in vivo. We confirmed the predictive value of ex vivo response to chemotherapy in a validation cohort of 95 individuals with acute myeloid leukemia treated with daunorubicin and cytarabine. Ex vivo drug response profiles improved ELN-22 risk stratification in individuals with adverse risk. We conclude that ex vivo drug response profiling is clinically feasible and has the potential to predict chemotherapy response in individuals with hematologic malignancies beyond clinically established genetic markers.

Systematic analysis of drug combinations against Gram-positive bacteria.

Drug combinations can expand options for antibacterial therapies but have not been systematically tested in Gram-positive species. We profiled ~8,000 combinations of 65 antibacterial drugs against the model species Bacillus subtilis and two prominent pathogens, Staphylococcus aureus and Streptococcus pneumoniae. Thereby, we recapitulated previously known drug interactions, but also identified ten times more novel interactions in the pathogen S. aureus, including 150 synergies. We showed that two synergies were equally effective against multidrug-resistant S. aureus clinical isolates in vitro and in vivo. Interactions were largely species-specific and synergies were distinct from those of Gram-negative species, owing to cell surface and drug uptake differences. We also tested 2,728 combinations of 44 commonly prescribed non-antibiotic drugs with 62 drugs with antibacterial activity against S. aureus and identified numerous antagonisms that might compromise the efficacy of antimicrobial therapies. We identified even more synergies and showed that the anti-aggregant ticagrelor synergized with cationic antibiotics by modifying the surface charge of S. aureus. All data can be browsed in an interactive interface ( https://apps.embl.de/combact/ ).

Comparing the value of mono- vs coculture for high-throughput compound screening in hematological malignancies.

Large-scale compound screens are a powerful model system for understanding variability of treatment response and discovering druggable tumor vulnerabilities of hematological malignancies. However, as mostly performed in a monoculture of tumor cells, these assays disregard modulatory effects of the in vivo microenvironment. It is an open question whether and to what extent coculture with bone marrow stromal cells could improve the biological relevance of drug testing assays over monoculture. Here, we established a high-throughput platform to measure ex vivo sensitivity of 108 primary blood cancer samples to 50 drugs in monoculture and coculture with bone marrow stromal cells. Stromal coculture conferred resistance to 52% of compounds in chronic lymphocytic leukemia (CLL) and 36% of compounds in acute myeloid leukemia (AML), including chemotherapeutics, B-cell receptor inhibitors, proteasome inhibitors, and Bromodomain and extraterminal domain inhibitors. Only the JAK inhibitors ruxolitinib and tofacitinib exhibited increased efficacy in AML and CLL stromal coculture. We further confirmed the importance of JAK-STAT signaling for stroma-mediated resistance by showing that stromal cells induce phosphorylation of STAT3 in CLL cells. We genetically characterized the 108 cancer samples and found that drug-gene associations strongly correlated between monoculture and coculture. However, effect sizes were lower in coculture, with more drug-gene associations detected in monoculture than in coculture. Our results justify a 2-step strategy for drug perturbation testing, with large-scale screening performed in monoculture, followed by focused evaluation of potential stroma-mediated resistances in coculture.

Subgroup-specific gene expression profiles and mixed epistasis in chronic lymphocytic leukemia.

Understanding the molecular and phenotypic heterogeneity of cancer is a prerequisite for effective treatment. For chronic lymphocytic leukemia (CLL), recurrent genetic driver events have been extensively cataloged, but this does not suffice to explain the disease's diverse course. Here, we performed RNA sequencing on 184 CLL patient samples. Unsupervised analysis revealed two major, orthogonal axes of gene expression variation: the first one represented the mutational status of the immunoglobulin heavy variable (IGHV) genes, and concomitantly, the three-group stratification of CLL by global DNA methylation. The second axis aligned with trisomy 12 status and affected chemokine, MAPK and mTOR signaling. We discovered non-additive effects (epistasis) of IGHV mutation status and trisomy 12 on multiple phenotypes, including the expression of 893 genes. Multiple types of epistasis were observed, including synergy, buffering, suppression and inversion, suggesting that molecular understanding of disease heterogeneity requires studying such genetic events not only individually but in combination. We detected strong differentially expressed gene signatures associated with major gene mutations and copy number aberrations including SF3B1, BRAF and TP53, as well as del(17)(p13), del(13)(q14) and del(11)(q22.3) beyond dosage effect. Our study reveals previously underappreciated gene expression signatures for the major molecular subtypes in CLL and the presence of epistasis between them.

Comparison of transformations for single-cell RNA-seq data.

The count table, a numeric matrix of genes × cells, is the basic input data structure in the analysis of single-cell RNA-sequencing data. A common preprocessing step is to adjust the counts for variable sampling efficiency and to transform them so that the variance is similar across the dynamic range. These steps are intended to make subsequent application of generic statistical methods more palatable. Here, we describe four transformation approaches based on the delta method, model residuals, inferred latent expression state and factor analysis. We compare their strengths and weaknesses and find that the latter three have appealing theoretical properties; however, in benchmarks using simulated and real-world data, it turns out that a rather simple approach, namely, the logarithm with a pseudo-count followed by principal-component analysis, performs as well or better than the more sophisticated alternatives. This result highlights limitations of current theoretical analysis as assessed by bottom-line performance benchmarks.

Deep thermal profiling for detection of functional proteoform groups.

The complexity of the functional proteome extends considerably beyond the coding genome, resulting in millions of proteoforms. Investigation of proteoforms and their functional roles is important to understand cellular physiology and its deregulation in diseases but challenging to perform systematically. Here we applied thermal proteome profiling with deep peptide coverage to detect functional proteoform groups in acute lymphoblastic leukemia cell lines with different cytogenetic aberrations. We detected 15,846 proteoforms, capturing differently spliced, cleaved and post-translationally modified proteins expressed from 9,290 genes. We identified differential co-aggregation of proteoform pairs and established links to disease biology. Moreover, we systematically made use of measured biophysical proteoform states to find specific biomarkers of drug sensitivity. Our approach, thus, provides a powerful and unique tool for systematic detection and functional annotation of proteoform groups.

htseq-clip: a toolset for the preprocessing of eCLIP/iCLIP datasets.

SUMMARY: Transcriptome-wide detection of binding sites of RNA-binding proteins is achieved using Individual-nucleotide crosslinking and immunoprecipitation (iCLIP) and its derivative enhanced CLIP (eCLIP) sequencing methods. Here, we introduce htseq-clip, a python package developed for preprocessing, extracting and summarizing crosslink site counts from i/eCLIP experimental data. The package delivers crosslink site count matrices along with other metrics, which can be directly used for filtering and downstream analyses such as the identification of differential binding sites. AVAILABILITY AND IMPLEMENTATION: The Python package htseq-clip is available via pypi (python package index), bioconda and the Galaxy Tool Shed under the open source MIT License. The code is hosted at https://github.com/EMBL-Hentze-group/htseq-clip and documentation is available under https://htseq-clip.readthedocs.io/en/latest.

Proteogenomics refines the molecular classification of chronic lymphocytic leukemia.

Cancer heterogeneity at the proteome level may explain differences in therapy response and prognosis beyond the currently established genomic and transcriptomic-based diagnostics. The relevance of proteomics for disease classifications remains to be established in clinically heterogeneous cancer entities such as chronic lymphocytic leukemia (CLL). Here, we characterize the proteome and transcriptome alongside genetic and ex-vivo drug response profiling in a clinically annotated CLL discovery cohort (n = 68). Unsupervised clustering of the proteome data reveals six subgroups. Five of these proteomic groups are associated with genetic features, while one group is only detectable at the proteome level. This new group is characterized by accelerated disease progression, high spliceosomal protein abundances associated with aberrant splicing, and low B cell receptor signaling protein abundances (ASB-CLL). Classifiers developed to identify ASB-CLL based on its characteristic proteome or splicing signature in two independent cohorts (n = 165, n = 169) confirm that ASB-CLL comprises about 20% of CLL patients. The inferior overall survival in ASB-CLL is also independent of both TP53- and IGHV mutation status. Our multi-omics analysis refines the classification of CLL and highlights the potential of proteomics to improve cancer patient stratification beyond genetic and transcriptomic profiling.

Drug-microenvironment perturbations reveal resistance mechanisms and prognostic subgroups in CLL.

The tumour microenvironment and genetic alterations collectively influence drug efficacy in cancer, but current evidence is limited and systematic analyses are lacking. Using chronic lymphocytic leukaemia (CLL) as a model disease, we investigated the influence of 17 microenvironmental stimuli on 12 drugs in 192 genetically characterised patient samples. Based on microenvironmental response, we identified four subgroups with distinct clinical outcomes beyond known prognostic markers. Response to multiple microenvironmental stimuli was amplified in trisomy 12 samples. Trisomy 12 was associated with a distinct epigenetic signature. Bromodomain inhibition reversed this epigenetic profile and could be used to target microenvironmental signalling in trisomy 12 CLL. We quantified the impact of microenvironmental stimuli on drug response and their dependence on genetic alterations, identifying interleukin 4 (IL4) and Toll-like receptor (TLR) stimulation as the strongest actuators of drug resistance. IL4 and TLR signalling activity was increased in CLL-infiltrated lymph nodes compared with healthy samples. High IL4 activity correlated with faster disease progression. The publicly available dataset can facilitate the investigation of cell-extrinsic mechanisms of drug resistance and disease progression.

Inferring tumor-specific cancer dependencies through integrating ex vivo drug response assays and drug-protein profiling.

The development of cancer therapies may be improved by the discovery of tumor-specific molecular dependencies. The requisite tools include genetic and chemical perturbations, each with its strengths and limitations. Chemical perturbations can be readily applied to primary cancer samples at large scale, but mechanistic understanding of hits and further pharmaceutical development is often complicated by the fact that a chemical compound has affinities to multiple proteins. To computationally infer specific molecular dependencies of individual cancers from their ex vivo drug sensitivity profiles, we developed a mathematical model that deconvolutes these data using measurements of protein-drug affinity profiles. Through integrating a drug-kinase profiling dataset and several drug response datasets, our method, DepInfeR, correctly identified known protein kinase dependencies, including the EGFR dependence of HER2+ breast cancer cell lines, the FLT3 dependence of acute myeloid leukemia (AML) with FLT3-ITD mutations and the differential dependencies on the B-cell receptor pathway in the two major subtypes of chronic lymphocytic leukemia (CLL). Furthermore, our method uncovered new subgroup-specific dependencies, including a previously unreported dependence of high-risk CLL on Checkpoint kinase 1 (CHEK1). The method also produced a detailed map of the kinase dependencies in a heterogeneous set of 117 CLL samples. The ability to deconvolute polypharmacological phenotypes into underlying causal molecular dependencies should increase the utility of high-throughput drug response assays for functional precision oncology.

Metabolic balance in colorectal cancer is maintained by optimal Wnt signaling levels.

Wnt pathways are important for the modulation of tissue homeostasis, and their deregulation is linked to cancer development. Canonical Wnt signaling is hyperactivated in many human colorectal cancers due to genetic alterations of the negative Wnt regulator APC. However, the expression levels of Wnt-dependent targets vary between tumors, and the mechanisms of carcinogenesis concomitant with this Wnt signaling dosage have not been understood. In this study, we integrate whole-genome CRISPR/Cas9 screens with large-scale multi-omic data to delineate functional subtypes of cancer. We engineer APC loss-of-function mutations and thereby hyperactivate Wnt signaling in cells with low endogenous Wnt activity and find that the resulting engineered cells have an unfavorable metabolic equilibrium compared with cells which naturally acquired Wnt hyperactivation. We show that the dosage level of oncogenic Wnt hyperactivation impacts the metabolic equilibrium and the mitochondrial phenotype of a given cell type in a context-dependent manner. These findings illustrate the impact of context-dependent genetic interactions on cellular phenotypes of a central cancer driver mutation and expand our understanding of quantitative modulation of oncogenic signaling in tumorigenesis.

Extravascular lung water levels are associated with mortality: a systematic review and meta-analysis.

BACKGROUND: The prognostic value of extravascular lung water (EVLW) measured by transpulmonary thermodilution (TPTD) in critically ill patients is debated. We performed a systematic review and meta-analysis of studies assessing the effects of TPTD-estimated EVLW on mortality in critically ill patients. METHODS: Cohort studies published in English from Embase, MEDLINE, and the Cochrane Database of Systematic Reviews from 1960 to 1 June 2021 were systematically searched. From eligible studies, the values of the odds ratio (OR) of EVLW as a risk factor for mortality, and the value of EVLW in survivors and non-survivors were extracted. Pooled OR were calculated from available studies. Mean differences and standard deviation of the EVLW between survivors and non-survivors were calculated. A random effects model was computed on the weighted mean differences across the two groups to estimate the pooled size effect. Subgroup analyses were performed to explore the possible sources of heterogeneity. RESULTS: Of the 18 studies included (1296 patients), OR could be extracted from 11 studies including 905 patients (464 survivors vs. 441 non-survivors), and 17 studies reported EVLW values of survivors and non-survivors, including 1246 patients (680 survivors vs. 566 non-survivors). The pooled OR of EVLW for mortality from eleven studies was 1.69 (95% confidence interval (CI) [1.22; 2.34], p < 0.0015). EVLW was significantly lower in survivors than non-survivors, with a mean difference of -4.97 mL/kg (95% CI [-6.54; -3.41], p < 0.001). The results regarding OR and mean differences were consistent in subgroup analyses. CONCLUSIONS: The value of EVLW measured by TPTD is associated with mortality in critically ill patients and is significantly higher in non-survivors than in survivors. This finding may also be interpreted as an indirect confirmation of the reliability of TPTD for estimating EVLW at the bedside. Nevertheless, our results should be considered cautiously due to the high risk of bias of many studies included in the meta-analysis and the low rating of certainty of evidence. Trial registration the study protocol was prospectively registered on PROSPERO: CRD42019126985.