Machine learning analysis of humoral and cellular responses to SARS-CoV-2 infection in young adults.

The severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) induces B and T cell responses, contributing to virus neutralization. In a cohort of 2,911 young adults, we identified 65 individuals who had an asymptomatic or mildly symptomatic SARS-CoV-2 infection and characterized their humoral and T cell responses to the Spike (S), Nucleocapsid (N) and Membrane (M) proteins. We found that previous infection induced CD4 T cells that vigorously responded to pools of peptides derived from the S and N proteins. By using statistical and machine learning models, we observed that the T cell response highly correlated with a compound titer of antibodies against the Receptor Binding Domain (RBD), S and N. However, while serum antibodies decayed over time, the cellular phenotype of these individuals remained stable over four months. Our computational analysis demonstrates that in young adults, asymptomatic and paucisymptomatic SARS-CoV-2 infections can induce robust and long-lasting CD4 T cell responses that exhibit slower decays than antibody titers. These observations imply that next-generation COVID-19 vaccines should be designed to induce stronger cellular responses to sustain the generation of potent neutralizing antibodies.

gExcite: a start-to-end framework for single-cell gene expression, hashing, and antibody analysis.

SUMMARY: Recently, CITE-seq emerged as a multimodal single-cell technology capturing gene expression and surface protein information from the same single cells, which allows unprecedented insights into disease mechanisms and heterogeneity, as well as immune cell profiling. Multiple single-cell profiling methods exist, but they are typically focused on either gene expression or antibody analysis, not their combination. Moreover, existing software suites are not easily scalable to a multitude of samples. To this end, we designed gExcite, a start-to-end workflow that provides both gene and antibody expression analysis, as well as hashing deconvolution. Embedded in the Snakemake workflow manager, gExcite facilitates reproducible and scalable analyses. We showcase the output of gExcite on a study of different dissociation protocols on PBMC samples. AVAILABILITY AND IMPLEMENTATION: gExcite is open source available on github at https://github.com/ETH-NEXUS/gExcite_pipeline. The software is distributed under the GNU General Public License 3 (GPL3).

Drug screening and genome editing in human pancreatic cancer organoids identifies drug-gene interactions and candidates for off-label treatment.

Pancreatic cancer (PDAC) is a highly aggressive malignancy for which the identification of novel therapies is urgently needed. Here, we establish a human PDAC organoid biobank from 31 genetically distinct lines, covering a representative range of tumor subtypes, and demonstrate that these reflect the molecular and phenotypic heterogeneity of primary PDAC tissue. We use CRISPR-Cas9 genome editing and drug screening to characterize drug-gene interactions with ARID1A and BRCA2. We find that missense- but not frameshift mutations in the PDAC driver gene ARID1A are associated with increased sensitivity to the kinase inhibitors dasatinib (p < 0.0001) and VE-821 (p < 0.0001). We conduct an automated drug-repurposing screen with 1,172 FDA-approved compounds, identifying 26 compounds that effectively kill PDAC organoids, including 19 chemotherapy drugs currently approved for other cancer types. We validate the activity of these compounds in vitro and in vivo. The in vivo validated hits include emetine and ouabain, compounds which are approved for non-cancer indications and which perturb the ability of PDAC organoids to respond to hypoxia. Our study provides proof-of-concept for advancing precision oncology and identifying candidates for drug repurposing via genome editing and drug screening in tumor organoid biobanks.

Impact of retrograde intrarenal surgery on biomarkers that are associated with renal parenchyma injury, a preliminary study.

PURPOSE: The primary objective of this preliminary study was to assess the changes in concentration of biomarkers, which indicate renal injury, after RIRS. MATERIALS AND METHODS: Within this prospective study, we included 21 patients with nephrolithiasis requiring treatment with RIRS. From each patient, blood and urine samples were taken at fixed intervals before and after RIRS. Kidney injury molecule-1 (KIM-1), monocyte chemoattractant protein 1 (MCP-1), neutrophil gelatinase-associated lipocalin (NGAL), interleukin-18 (IL-18), calbindin, albumin, clusterin, gluthation S-transferase-π (GST-π), beta-2-microglobulin (B2M), osteopontin, cystatin c, and trefoil-factor-3 (TFF3) were measured in urine. Creatinine, cystatin c and uric acid were analyzed in the blood samples. RESULTS: A significant increase of the biomarkers clusterin, GST-π, B2M, NGAL and cystatin c was observed after RIRS. However, the biomarkers gradually normalized during the first 14 postoperative days. The parameters surgery time, cumulative stone volume, and BMI did not significantly influence the biomarker concentrations. In the case of GST-π and NGAL a significant positive, yet minuscule effect of age was observed. CONCLUSIONS: With our study, we identified 5 out of 12 assessed renal injury biomarkers that showed a significant increase after RIRS. The increase was only temporary and all markers normalized within 14 days. Further studies are needed to determine the clinical value of these identified markers to assess the long-term impact of intrarenal pressure elevation during RIRS.

A showcase study on personalized in silico drug response prediction based on the genetic landscape of muscle invasive bladder cancer.

Improved and cheaper molecular diagnostics allow the shift from "one size fits all" therapies to personalised treatments targeting the individual tumor. However, the wealth of potential targets based on comprehensive sequencing remains a yet unsolved challenge that prevents its routine use in clinical practice. Thus, we designed a workflow that selects the most promising treatment targets based on multi-omics sequencing and in silico drug prediction. In this study we demonstrate the workflow with focus on bladder cancer (BLCA), as there are, to date, no reliable diagnostics available to predict the potential benefit of a therapeutic approach. Within the TCGA-BLCA cohort, our workflow identified a panel of 21 genes and 72 drugs that suggested personalized treatment for 95% of patients-including five genes not yet reported as prognostic markers for clinical testing in BLCA. The automated predictions were complemented by manually curated data, thus allowing for accurate sensitivity- or resistance-directed drug response predictions. We discuss potential improvements of drug-gene interaction databases on the basis of pitfalls that were identified during manual curation.

The Tumor Profiler Study: integrated, multi-omic, functional tumor profiling for clinical decision support.

The application and integration of molecular profiling technologies create novel opportunities for personalized medicine. Here, we introduce the Tumor Profiler Study, an observational trial combining a prospective diagnostic approach to assess the relevance of in-depth tumor profiling to support clinical decision-making with an exploratory approach to improve the biological understanding of the disease.

Text-Mining Services of the Swiss Variant Interpretation Platform for Oncology.

The Swiss Variant Interpretation Platform for Oncology is a centralized, joint and curated database for clinical somatic variants piloted by a board of Swiss healthcare institutions and operated by the SIB Swiss Institute of Bioinformatics. To support this effort, SIB Text Mining designed a set of text analytics services. This report focuses on three of those services. First, the automatic annotations of the literature with a set of terminologies have been performed, resulting in a large annotated version of MEDLINE and PMC. Second, a generator of variant synonyms for single nucleotide variants has been developed using publicly available data resources, as well as patterns of non-standard formats, often found in the literature. Third, a literature ranking service enables to retrieve a ranked set of MEDLINE abstracts given a variant and optionally a diagnosis. The annotation of MEDLINE and PMC resulted in a total of respectively 785,181,199 and 1,156,060,212 annotations, which means an average of 26 and 425 annotations per abstract and full-text article. The generator of variant synonyms enables to retrieve up to 42 synonyms for a variant. The literature ranking service reaches a precision (P10) of 63%, which means that almost two-thirds of the top-10 returned abstracts are judged relevant. Further services will be implemented to complete this set of services, such as a service to retrieve relevant clinical trials for a patient and a literature ranking service for full-text articles.

Lower urinary tract symptoms (LUTS) before and after robotic-assisted laparoscopic prostatectomy: does improvement of LUTS mitigate worsened incontinence after robotic prostatectomy?

BACKGROUND: Urinary incontinence is a major concern for patients scheduled for radical prostatectomy. However, after prostatectomy lower urinary tract symptoms (LUTS) may improve and thus mitigate this concern. We assessed LUTS and its interference with the quality of life (QoL) using the short form of the international continence society male questionnaire (ICSMALESF-Q) in patients before and after robot-assisted radical prostatectomy (RARP). Furthermore, we aimed to identify risk factors for postoperative urinary incontinence. METHODS: Data of all patients who underwent RARP from 2009 to 2014 were prospectively collected in our customized database. We identified 453 eligible patients for whom a preoperative and at least two postoperative datasets including ICSMALESF-Q were available. RESULTS: Both the ICSMALESF-Q at 6 months (P<0.001) and the related QoL at 12 months (P<0.01) have significantly improved after RARP (P<0.001). Two years after RARP ICSMALESF-Q and thus LUTS have improved in 64%, remained unchanged in 18% and worsened in 18% of patients. The daily pad use was 0 in 79% and 0 or 1 pad in 95.6%, respectively. Increased patient age (P<0.05) was significantly associated with an increased average number of pads used per day (multiplicative effect: +2.1% pads for each year). Being in the D'Amico low-risk group reduced the average number of pads used by 22% (P<0.05, multiplicative effect 0.780). The prostate volume, planned nerve sparing, adjuvant or salvage radiotherapy, body mass index (BMI), or a history of transurethral resection of the prostate (TUR-P) before radical prostatectomy were not associated with the postoperative pad use or changes in LUTS. CONCLUSIONS: The ICSMALESF-Q and thus LUTS have significantly improved in a majority of patients after RARP and hence the associated QoL improved as well. Preoperative D'Amico low-risk group significantly reduced pad use after RARP, whereas increased age significantly increased postoperative pad use. These results will help providers counsel their patients more appropriately before prostatectomy by focusing not only on pad use and incontinence after RARP, but also on changes of the bothersomeness of LUTS and risk factors in general.

Bioinformatics for precision oncology.

Molecular profiling of tumor biopsies plays an increasingly important role not only in cancer research, but also in the clinical management of cancer patients. Multi-omics approaches hold the promise of improving diagnostics, prognostics and personalized treatment. To deliver on this promise of precision oncology, appropriate bioinformatics methods for managing, integrating and analyzing large and complex data are necessary. Here, we discuss the specific requirements of bioinformatics methods and software that arise in the setting of clinical oncology, owing to a stricter regulatory environment and the need for rapid, highly reproducible and robust procedures. We describe the workflow of a molecular tumor board and the specific bioinformatics support that it requires, from the primary analysis of raw molecular profiling data to the automatic generation of a clinical report and its delivery to decision-making clinical oncologists. Such workflows have to various degrees been implemented in many clinical trials, as well as in molecular tumor boards at specialized cancer centers and university hospitals worldwide. We review these and more recent efforts to include other high-dimensional multi-omics patient profiles into the tumor board, as well as the state of clinical decision support software to translate molecular findings into treatment recommendations.

SwissMTB: establishing comprehensive molecular cancer diagnostics in Swiss clinics.

BACKGROUND: Molecular precision oncology is an emerging practice to improve cancer therapy by decreasing the risk of choosing treatments that lack efficacy or cause adverse events. However, the challenges of integrating molecular profiling into routine clinical care are manifold. From a computational perspective these include the importance of a short analysis turnaround time, the interpretation of complex drug-gene and gene-gene interactions, and the necessity of standardized high-quality workflows. In addition, difficulties faced when integrating molecular diagnostics into clinical practice are ethical concerns, legal requirements, and limited availability of treatment options beyond standard of care as well as the overall lack of awareness of their existence. METHODS: To the best of our knowledge, we are the first group in Switzerland that established a workflow for personalized diagnostics based on comprehensive high-throughput sequencing of tumors at the clinic. Our workflow, named SwissMTB (Swiss Molecular Tumor Board), links genetic tumor alterations and gene expression to therapeutic options and clinical trial opportunities. The resulting treatment recommendations are summarized in a clinical report and discussed in a molecular tumor board at the clinic to support therapy decisions. RESULTS: Here we present results from an observational pilot study including 22 late-stage cancer patients. In this study we were able to identify actionable variants and corresponding therapies for 19 patients. Half of the patients were analyzed retrospectively. In two patients we identified resistance-associated variants explaining lack of therapy response. For five out of eleven patients analyzed before treatment the SwissMTB diagnostic influenced treatment decision. CONCLUSIONS: SwissMTB enables the analysis and clinical interpretation of large numbers of potentially actionable molecular targets. Thus, our workflow paves the way towards a more frequent use of comprehensive molecular diagnostics in Swiss hospitals.

Linear Mixed Effects Analysis Reveals the Significant Impact of Preoperative Diet Success on Postoperative Weight Loss in Gastric Bypass Surgery.

BACKGROUND: Several studies investigated the impact of preoperative weight loss on bariatric surgery outcome. However, they mostly focus on small groups of patients or lack updated statistical support. METHODS: Two hundred and thirty-nine consecutive patients undergoing laparoscopic, proximal Roux-en-Y gastric bypass at our institution between September 2009 and November 2015 were studied. Patients were operated by the same surgeon, applying a standardized technique and followed a 500-kcal/day preoperative diet, starting 14 days before surgery. Body weight was measured before diet, at surgery, and at least three times postoperatively. A linear mixed effects (LME) model and Benedict and Harris formula were used to assess association of pre- and postoperative weight loss up to 2 years postoperatively. RESULTS: Patients' (184 females) initial weight was 121.7 kg (females 117.2 kg; males 136.6 kg). They lost on average 5.3 kg (females 4.7 kg; males 7.0 kg) pre- and 36.8 kg (females 36.7 kg; males 37.0 kg) postoperatively, within 2 years. Average excess weight loss (EWL) was 67.2% (females 66.6%; males 67.4%). In 205 patients (154 females), EWL exceeded 50%. Longitudinal data analysis according to LME showed a significant impact of pre- on postoperative weight loss (p < 0.001, likelihood-ratio test, LRT). These effects were undetectable if patients were evaluated by non-parametric analysis based on application of the Benedict and Harris formula. CONCLUSIONS: Preoperative dietary success is associated with postoperative weight loss. Effects predicted by the LME model are most pronounced in the first 4-6 months after surgery and are fading away within 24 months postoperatively. External factors not considered in this study might dominate in later phases.

NGS-pipe: a flexible, easily extendable and highly configurable framework for NGS analysis.

Motivation: Next-generation sequencing is now an established method in genomics, and massive amounts of sequencing data are being generated on a regular basis. Analysis of the sequencing data is typically performed by lab-specific in-house solutions, but the agreement of results from different facilities is often small. General standards for quality control, reproducibility and documentation are missing. Results: We developed NGS-pipe, a flexible, transparent and easy-to-use framework for the design of pipelines to analyze whole-exome, whole-genome and transcriptome sequencing data. NGS-pipe facilitates the harmonization of genomic data analysis by supporting quality control, documentation, reproducibility, parallelization and easy adaptation to other NGS experiments. Availability and implementation: https://github.com/cbg-ethz/NGS-pipe. Contact: niko.beerenwinkel@bsse.ethz.ch.

Genomic variant annotation workflow for clinical applications.

Annotation and interpretation of DNA aberrations identified through next-generation sequencing is becoming an increasingly important task. Even more so in the context of data analysis pipelines for medical applications, where genomic aberrations are associated with phenotypic and clinical features. Here we describe a workflow to identify potential gene targets in aberrated genes or pathways and their corresponding drugs. To this end, we provide the R/Bioconductor package rDGIdb, an R wrapper to query the drug-gene interaction database (DGIdb). DGIdb accumulates drug-gene interaction data from 15 different resources and allows filtering on different levels. The rDGIdb package makes these resources and tools available to R users. Moreover, rDGIdb queries can be automated through incorporation of the rDGIdb package into NGS sequencing pipelines.

Improved prediction of peptide detectability for targeted proteomics using a rank-based algorithm and organism-specific data.

The in silico prediction of the best-observable "proteotypic" peptides in mass spectrometry-based workflows is a challenging problem. Being able to accurately predict such peptides would enable the informed selection of proteotypic peptides for targeted quantification of previously observed and non-observed proteins for any organism, with a significant impact for clinical proteomics and systems biology studies. Current prediction algorithms rely on physicochemical parameters in combination with positive and negative training sets to identify those peptide properties that most profoundly affect their general detectability. Here we present PeptideRank, an approach that uses learning to rank algorithm for peptide detectability prediction from shotgun proteomics data, and that eliminates the need to select a negative dataset for the training step. A large number of different peptide properties are used to train ranking models in order to predict a ranking of the best-observable peptides within a protein. Empirical evaluation with rank accuracy metrics showed that PeptideRank complements existing prediction algorithms. Our results indicate that the best performance is achieved when it is trained on organism-specific shotgun proteomics data, and that PeptideRank is most accurate for short to medium-sized and abundant proteins, without any loss in prediction accuracy for the important class of membrane proteins. BIOLOGICAL SIGNIFICANCE: Targeted proteomics approaches have been gaining a lot of momentum and hold immense potential for systems biology studies and clinical proteomics. However, since only very few complete proteomes have been reported to date, for a considerable fraction of a proteome there is no experimental proteomics evidence that would allow to guide the selection of the best-suited proteotypic peptides (PTPs), i.e. peptides that are specific to a given proteoform and that are repeatedly observed in a mass spectrometer. We describe a novel, rank-based approach for the prediction of the best-suited PTPs for targeted proteomics applications. By building on methods developed in the field of information retrieval (e.g. web search engines like Google's PageRank), we circumvent the delicate step of selecting positive and negative training sets and at the same time also more closely reflect the experimentalist´s need for selecting e.g. the 5 most promising peptides for targeting a protein of interest. This approach allows to predict PTPs for not yet observed proteins or for organisms without prior experimental proteomics data such as many non-model organisms.

Proteome-wide identification of predominant subcellular protein localizations in a bacterial model organism.

Proteomics data provide unique insights into biological systems, including the predominant subcellular localization (SCL) of proteins, which can reveal important clues about their functions. Here we analyzed data of a complete prokaryotic proteome expressed under two conditions mimicking interaction of the emerging pathogen Bartonella henselae with its mammalian host. Normalized spectral count data from cytoplasmic, total membrane, inner and outer membrane fractions allowed us to identify the predominant SCL for 82% of the identified proteins. The spectral count proportion of total membrane versus cytoplasmic fractions indicated the propensity of cytoplasmic proteins to co-fractionate with the inner membrane, and enabled us to distinguish cytoplasmic, peripheral inner membrane and bona fide inner membrane proteins. Principal component analysis and k-nearest neighbor classification training on selected marker proteins or predominantly localized proteins, allowed us to determine an extensive catalog of at least 74 expressed outer membrane proteins, and to extend the SCL assignment to 94% of the identified proteins, including 18% where in silico methods gave no prediction. Suitable experimental proteomics data combined with straightforward computational approaches can thus identify the predominant SCL on a proteome-wide scale. Finally, we present a conceptual approach to identify proteins potentially changing their SCL in a condition-dependent fashion. BIOLOGICAL SIGNIFICANCE: The work presented here describes the first prokaryotic proteome-wide subcellular localization (SCL) dataset for the emerging pathogen B. henselae (Bhen). The study indicates that suitable subcellular fractionation experiments combined with straight-forward computational analysis approaches assessing the proportion of spectral counts observed in different subcellular fractions are powerful for determining the predominant SCL of a large percentage of the experimentally observed proteins. This includes numerous cases where in silico prediction methods do not provide any prediction. Avoiding a treatment with harsh conditions, cytoplasmic proteins tend to co-fractionate with proteins of the inner membrane fraction, indicative of close functional interactions. The spectral count proportion (SCP) of total membrane versus cytoplasmic fractions allowed us to obtain a good indication about the relative proximity of individual protein complex members to the inner membrane. Using principal component analysis and k-nearest neighbor approaches, we were able to extend the percentage of proteins with a predominant experimental localization to over 90% of all expressed proteins and identified a set of at least 74 outer membrane (OM) proteins. In general, OM proteins represent a rich source of candidates for the development of urgently needed new therapeutics in combat of resurgence of infectious disease and multi-drug resistant bacteria. Finally, by comparing the data from two infection biology relevant conditions, we conceptually explore methods to identify and visualize potential candidates that may partially change their SCL in these different conditions. The data are made available to researchers as a SCL compendium for Bhen and as an assistance in further improving in silico SCL prediction algorithms.

Directed shotgun proteomics guided by saturated RNA-seq identifies a complete expressed prokaryotic proteome.

Prokaryotes, due to their moderate complexity, are particularly amenable to the comprehensive identification of the protein repertoire expressed under different conditions. We applied a generic strategy to identify a complete expressed prokaryotic proteome, which is based on the analysis of RNA and proteins extracted from matched samples. Saturated transcriptome profiling by RNA-seq provided an endpoint estimate of the protein-coding genes expressed under two conditions which mimic the interaction of Bartonella henselae with its mammalian host. Directed shotgun proteomics experiments were carried out on four subcellular fractions. By specifically targeting proteins which are short, basic, low abundant, and membrane localized, we could eliminate their initial underrepresentation compared to the estimated endpoint. A total of 1250 proteins were identified with an estimated false discovery rate below 1%. This represents 85% of all distinct annotated proteins and ∼90% of the expressed protein-coding genes. Genes that were detected at the transcript but not protein level, were found to be highly enriched in several genomic islands. Furthermore, genes that lacked an ortholog and a functional annotation were not detected at the protein level; these may represent examples of overprediction in genome annotations. A dramatic membrane proteome reorganization was observed, including differential regulation of autotransporters, adhesins, and hemin binding proteins. Particularly noteworthy was the complete membrane proteome coverage, which included expression of all members of the VirB/D4 type IV secretion system, a key virulence factor.

Causal stability ranking.

Genotypic causes of a phenotypic trait are typically determined via randomized controlled intervention experiments. Such experiments are often prohibitive with respect to durations and costs, and informative prioritization of experiments is desirable. We therefore consider predicting stable rankings of genes (covariates), according to their total causal effects on a phenotype (response), from observational data. Since causal effects are generally non-identifiable from observational data only, we use a method that can infer lower bounds for the total causal effect under some assumptions. We validated our method, which we call Causal Stability Ranking (CStaR), in two situations. First, we performed knock-out experiments with Arabidopsis thaliana according to a predicted ranking based on observational gene expression data, using flowering time as phenotype of interest. Besides several known regulators of flowering time, we found almost half of the tested top ranking mutants to have a significantly changed flowering time. Second, we compared CStaR to established regression-based methods on a gene expression dataset of Saccharomyces cerevisiae. We found that CStaR outperforms these established methods. Our method allows for efficient design and prioritization of future intervention experiments, and due to its generality it can be used for a broad spectrum of applications.

MissForest--non-parametric missing value imputation for mixed-type data.

MOTIVATION: Modern data acquisition based on high-throughput technology is often facing the problem of missing data. Algorithms commonly used in the analysis of such large-scale data often depend on a complete set. Missing value imputation offers a solution to this problem. However, the majority of available imputation methods are restricted to one type of variable only: continuous or categorical. For mixed-type data, the different types are usually handled separately. Therefore, these methods ignore possible relations between variable types. We propose a non-parametric method which can cope with different types of variables simultaneously. RESULTS: We compare several state of the art methods for the imputation of missing values. We propose and evaluate an iterative imputation method (missForest) based on a random forest. By averaging over many unpruned classification or regression trees, random forest intrinsically constitutes a multiple imputation scheme. Using the built-in out-of-bag error estimates of random forest, we are able to estimate the imputation error without the need of a test set. Evaluation is performed on multiple datasets coming from a diverse selection of biological fields with artificially introduced missing values ranging from 10% to 30%. We show that missForest can successfully handle missing values, particularly in datasets including different types of variables. In our comparative study, missForest outperforms other methods of imputation especially in data settings where complex interactions and non-linear relations are suspected. The out-of-bag imputation error estimates of missForest prove to be adequate in all settings. Additionally, missForest exhibits attractive computational efficiency and can cope with high-dimensional data. AVAILABILITY: The package missForest is freely available from http://stat.ethz.ch/CRAN/. CONTACT: stekhoven@stat.math.ethz.ch; buhlmann@stat.math.ethz.ch