Predicting genes associated with RNA methylation pathways using machine learning.

RNA methylation plays an important role in functional regulation of RNAs, and has thus attracted an increasing interest in biology and drug discovery. Here, we collected and collated transcriptomic, proteomic, structural and physical interaction data from the Harmonizome database, and applied supervised machine learning to predict novel genes associated with RNA methylation pathways in human. We selected five types of classifiers, which we trained and evaluated using cross-validation on multiple training sets. The best models reached 88% accuracy based on cross-validation, and an average 91% accuracy on the test set. Using protein-protein interaction data, we propose six molecular sub-networks linking model predictions to previously known RNA methylation genes, with roles in mRNA methylation, tRNA processing, rRNA processing, but also protein and chromatin modifications. Our study exemplifies how access to large omics datasets joined by machine learning methods can be used to predict gene function.

Novel Insights into Quantitative Proteomics from an Innovative Bottom-Up Simple Light Isotope Metabolic (bSLIM) Labeling Data Processing Strategy.

Simple light isotope metabolic labeling (SLIM labeling) is an innovative method to quantify variations in the proteome based on an original in vivo labeling strategy. Heterotrophic cells grown in U-[12C] as the sole source of carbon synthesize U-[12C]-amino acids, which are incorporated into proteins, giving rise to U-[12C]-proteins. This results in a large increase in the intensity of the monoisotope ion of peptides and proteins, thus allowing higher identification scores and protein sequence coverage in mass spectrometry experiments. This method, initially developed for signal processing and quantification of the incorporation rate of 12C into peptides, was based on a multistep process that was difficult to implement for many laboratories. To overcome these limitations, we developed a new theoretical background to analyze bottom-up proteomics data using SLIM-labeling (bSLIM) and established simple procedures based on open-source software, using dedicated OpenMS modules, and embedded R scripts to process the bSLIM experimental data. These new tools allow computation of both the 12C abundance in peptides to follow the kinetics of protein labeling and the molar fraction of unlabeled and 12C-labeled peptides in multiplexing experiments to determine the relative abundance of proteins extracted under different biological conditions. They also make it possible to consider incomplete 12C labeling, such as that observed in cells with nutritional requirements for nonlabeled amino acids. These tools were validated on an experimental dataset produced using various yeast strains of Saccharomyces cerevisiae and growth conditions. The workflows are built on the implementation of appropriate calculation modules in a KNIME working environment. These new integrated tools provide a convenient framework for the wider use of the SLIM-labeling strategy.

Phosphorylation-Dependent Assembly of a 14-3-3 Mediated Signaling Complex during Red Blood Cell Invasion by Plasmodium falciparum Merozoites.

Red blood cell (RBC) invasion by Plasmodium merozoites requires multiple steps that are regulated by signaling pathways. Exposure of P. falciparum merozoites to the physiological signal of low K+, as found in blood plasma, leads to a rise in cytosolic Ca2+, which mediates microneme secretion, motility, and invasion. We have used global phosphoproteomic analysis of merozoites to identify signaling pathways that are activated during invasion. Using quantitative phosphoproteomics, we found 394 protein phosphorylation site changes in merozoites subjected to different ionic environments (high K+/low K+), 143 of which were Ca2+ dependent. These included a number of signaling proteins such as catalytic and regulatory subunits of protein kinase A (PfPKAc and PfPKAr) and calcium-dependent protein kinase 1 (PfCDPK1). Proteins of the 14-3-3 family interact with phosphorylated target proteins to assemble signaling complexes. Here, using coimmunoprecipitation and gel filtration chromatography, we demonstrate that Pf14-3-3I binds phosphorylated PfPKAr and PfCDPK1 to mediate the assembly of a multiprotein complex in P. falciparum merozoites. A phospho-peptide, P1, based on the Ca2+-dependent phosphosites of PKAr, binds Pf14-3-3I and disrupts assembly of the Pf14-3-3I-mediated multiprotein complex. Disruption of the multiprotein complex with P1 inhibits microneme secretion and RBC invasion. This study thus identifies a novel signaling complex that plays a key role in merozoite invasion of RBCs. Disruption of this signaling complex could serve as a novel approach to inhibit blood-stage growth of malaria parasites.IMPORTANCE Invasion of red blood cells (RBCs) by Plasmodium falciparum merozoites is a complex process that is regulated by intricate signaling pathways. Here, we used phosphoproteomic profiling to identify the key proteins involved in signaling events during invasion. We found changes in the phosphorylation of various merozoite proteins, including multiple kinases previously implicated in the process of invasion. We also found that a phosphorylation-dependent multiprotein complex including signaling kinases assembles during the process of invasion. Disruption of this multiprotein complex impairs merozoite invasion of RBCs, providing a novel approach for the development of inhibitors to block the growth of blood-stage malaria parasites.

Genome-wide investigation of gene-cancer associations for the prediction of novel therapeutic targets in oncology.

A major cause of failed drug discovery programs is suboptimal target selection, resulting in the development of drug candidates that are potent inhibitors, but ineffective at treating the disease. In the genomics era, the availability of large biomedical datasets with genome-wide readouts has the potential to transform target selection and validation. In this study we investigate how computational intelligence methods can be applied to predict novel therapeutic targets in oncology. We compared different machine learning classifiers applied to the task of drug target classification for nine different human cancer types. For each cancer type, a set of "known" target genes was obtained and equally-sized sets of "non-targets" were sampled multiple times from the human protein-coding genes. Models were trained on mutation, gene expression (TCGA), and gene essentiality (DepMap) data. In addition, we generated a numerical embedding of the interaction network of protein-coding genes using deep network representation learning and included the results in the modeling. We assessed feature importance using a random forests classifier and performed feature selection based on measuring permutation importance against a null distribution. Our best models achieved good generalization performance based on the AUROC metric. With the best model for each cancer type, we ran predictions on more than 15,000 protein-coding genes to identify potential novel targets. Our results indicate that this approach may be useful to inform early stages of the drug discovery pipeline.

A computational platform for high-throughput analysis of RNA sequences and modifications by mass spectrometry.

The field of epitranscriptomics continues to reveal how post-transcriptional modification of RNA affects a wide variety of biological phenomena. A pivotal challenge in this area is the identification of modified RNA residues within their sequence contexts. Mass spectrometry (MS) offers a comprehensive solution by using analogous approaches to shotgun proteomics. However, software support for the analysis of RNA MS data is inadequate at present and does not allow high-throughput processing. Existing software solutions lack the raw performance and statistical grounding to efficiently handle the numerous modifications found on RNA. We present a free and open-source database search engine for RNA MS data, called NucleicAcidSearchEngine (NASE), that addresses these shortcomings. We demonstrate the capability of NASE to reliably identify a wide range of modified RNA sequences in four original datasets of varying complexity. In human tRNA, we characterize over 20 different modification types simultaneously and find many cases of incomplete modification.

Evaluation of a Dual Isolation Width Acquisition Method for Isobaric Labeling Ratio Decompression.

Isobaric labeling is a highly precise approach for protein quantification. However, due to the isolation interference problem, isobaric tagging suffers from ratio underestimation at the MS2 level. The use of narrow isolation widths is a rational approach to alleviate the interference problem; however, this approach compromises proteome coverage. We reasoned that although a very narrow isolation window will result in loss of peptide fragment ions, the reporter ion signals will be retained for a significant portion of the spectra. On the basis of this assumption, we have designed a dual isolation width acquisition (DIWA) method, in which each precursor is first fragmented with HCD using a standard isolation width for peptide identification and preliminary quantification, followed by a second MS2 HCD scan using a much narrower isolation width for the acquisition of quantitative spectra with reduced interference. We leverage the quantification obtained by the "narrow" scans to build linear regression models and apply these to decompress the fold-changes measured at the "standard" scans. We evaluate the DIWA approach using a nested two species/gene knockout TMT-6plex experimental design and discuss the perspectives of this approach.

Targeted Feature Detection for Data-Dependent Shotgun Proteomics.

Label-free quantification of shotgun LC-MS/MS data is the prevailing approach in quantitative proteomics but remains computationally nontrivial. The central data analysis step is the detection of peptide-specific signal patterns, called features. Peptide quantification is facilitated by associating signal intensities in features with peptide sequences derived from MS2 spectra; however, missing values due to imperfect feature detection are a common problem. A feature detection approach that directly targets identified peptides (minimizing missing values) but also offers robustness against false-positive features (by assigning meaningful confidence scores) would thus be highly desirable. We developed a new feature detection algorithm within the OpenMS software framework, leveraging ideas and algorithms from the OpenSWATH toolset for DIA/SRM data analysis. Our software, FeatureFinderIdentification ("FFId"), implements a targeted approach to feature detection based on information from identified peptides. This information is encoded in an MS1 assay library, based on which ion chromatogram extraction and detection of feature candidates are carried out. Significantly, when analyzing data from experiments comprising multiple samples, our approach distinguishes between "internal" and "external" (inferred) peptide identifications (IDs) for each sample. On the basis of internal IDs, two sets of positive (true) and negative (decoy) feature candidates are defined. A support vector machine (SVM) classifier is then trained to discriminate between the sets and is subsequently applied to the "uncertain" feature candidates from external IDs, facilitating selection and confidence scoring of the best feature candidate for each peptide. This approach also enables our algorithm to estimate the false discovery rate (FDR) of the feature selection step. We validated FFId based on a public benchmark data set, comprising a yeast cell lysate spiked with protein standards that provide a known ground-truth. The algorithm reached almost complete (>99%) quantification coverage for the full set of peptides identified at 1% FDR (PSM level). Compared with other software solutions for label-free quantification, this is an outstanding result, which was achieved at competitive quantification accuracy and reproducibility across replicates. The FDR for the feature selection was estimated at a low 1.5% on average per sample (3% for features inferred from external peptide IDs). The FFId software is open-source and freely available as part of OpenMS ( www.openms.org ).

Flexible Data Analysis Pipeline for High-Confidence Proteogenomics.

Proteogenomics leverages information derived from proteomic data to improve genome annotations. Of particular interest are "novel" peptides that provide direct evidence of protein expression for genomic regions not previously annotated as protein-coding. We present a modular, automated data analysis pipeline aimed at detecting such "novel" peptides in proteomic data sets. This pipeline implements criteria developed by proteomics and genome annotation experts for high-stringency peptide identification and filtering. Our pipeline is based on the OpenMS computational framework; it incorporates multiple database search engines for peptide identification and applies a machine-learning approach (Percolator) to post-process search results. We describe several new and improved software tools that we developed to facilitate proteogenomic analyses that enhance the wealth of tools provided by OpenMS. We demonstrate the application of our pipeline to a human testis tissue data set previously acquired for the Chromosome-Centric Human Proteome Project, which led to the addition of five new gene annotations on the human reference genome.

OpenMS: a flexible open-source software platform for mass spectrometry data analysis.

High-resolution mass spectrometry (MS) has become an important tool in the life sciences, contributing to the diagnosis and understanding of human diseases, elucidating biomolecular structural information and characterizing cellular signaling networks. However, the rapid growth in the volume and complexity of MS data makes transparent, accurate and reproducible analysis difficult. We present OpenMS 2.0 (http://www.openms.de), a robust, open-source, cross-platform software specifically designed for the flexible and reproducible analysis of high-throughput MS data. The extensible OpenMS software implements common mass spectrometric data processing tasks through a well-defined application programming interface in C++ and Python and through standardized open data formats. OpenMS additionally provides a set of 185 tools and ready-made workflows for common mass spectrometric data processing tasks, which enable users to perform complex quantitative mass spectrometric analyses with ease.

Improving GENCODE reference gene annotation using a high-stringency proteogenomics workflow.

Complete annotation of the human genome is indispensable for medical research. The GENCODE consortium strives to provide this, augmenting computational and experimental evidence with manual annotation. The rapidly developing field of proteogenomics provides evidence for the translation of genes into proteins and can be used to discover and refine gene models. However, for both the proteomics and annotation groups, there is a lack of guidelines for integrating this data. Here we report a stringent workflow for the interpretation of proteogenomic data that could be used by the annotation community to interpret novel proteogenomic evidence. Based on reprocessing of three large-scale publicly available human data sets, we show that a conservative approach, using stringent filtering is required to generate valid identifications. Evidence has been found supporting 16 novel protein-coding genes being added to GENCODE. Despite this many peptide identifications in pseudogenes cannot be annotated due to the absence of orthogonal supporting evidence.

Expression Atlas update--an integrated database of gene and protein expression in humans, animals and plants.

Expression Atlas (http://www.ebi.ac.uk/gxa) provides information about gene and protein expression in animal and plant samples of different cell types, organism parts, developmental stages, diseases and other conditions. It consists of selected microarray and RNA-sequencing studies from ArrayExpress, which have been manually curated, annotated with ontology terms, checked for high quality and processed using standardised analysis methods. Since the last update, Atlas has grown seven-fold (1572 studies as of August 2015), and incorporates baseline expression profiles of tissues from Human Protein Atlas, GTEx and FANTOM5, and of cancer cell lines from ENCODE, CCLE and Genentech projects. Plant studies constitute a quarter of Atlas data. For genes of interest, the user can view baseline expression in tissues, and differential expression for biologically meaningful pairwise comparisons-estimated using consistent methodology across all of Atlas. Our first proteomics study in human tissues is now displayed alongside transcriptomics data in the same tissues. Novel analyses and visualisations include: 'enrichment' in each differential comparison of GO terms, Reactome, Plant Reactome pathways and InterPro domains; hierarchical clustering (by baseline expression) of most variable genes and experimental conditions; and, for a given gene-condition, distribution of baseline expression across biological replicates.

Prediction of colorectal cancer diagnosis based on circulating plasma proteins.

Non-invasive detection of colorectal cancer with blood-based markers is a critical clinical need. Here we describe a phased mass spectrometry-based approach for the discovery, screening, and validation of circulating protein biomarkers with diagnostic value. Initially, we profiled human primary tumor tissue epithelia and characterized about 300 secreted and cell surface candidate glycoproteins. These candidates were then screened in patient systemic circulation to identify detectable candidates in blood plasma. An 88-plex targeting method was established to systematically monitor these proteins in two large and independent cohorts of plasma samples, which generated quantitative clinical datasets at an unprecedented scale. The data were deployed to develop and evaluate a five-protein biomarker signature for colorectal cancer detection.

Dynamic phosphoproteomics reveals TORC1-dependent regulation of yeast nucleotide and amino acid biosynthesis.

Phosphoproteomics studies have unraveled the extent of protein phosphorylation as a key cellular regulation mechanism, but assigning functionality to specific phosphorylation events remains a major challenge. TORC1 (target of rapamycin complex 1) is a kinase-containing protein complex that transduces changes in nutrient availability into phosphorylation signaling events that alter cell growth and proliferation. To resolve the temporal sequence of phosphorylation responses to nutritionally and chemically induced changes in TORC1 signaling and to identify previously unknown kinase-substrate relationships in Saccharomyces cerevisiae, we performed quantitative mass spectrometry-based phosphoproteomic analyses after shifts in nitrogen sources and rapamycin treatment. From early phosphorylation events that were consistent over at least two experimental perturbations, we identified 51 candidate and 10 known proximal targets of TORC1 that were direct substrates of TORC1 or of one of its kinase or phosphatase substrates. By correlating these phosphoproteomics data with dynamic metabolomics data, we inferred the functional role of phosphorylation on the metabolic activity of 12 enzymes, including three candidate TORC1-proximal targets: Amd1, which is involved in nucleotide metabolism; Hom3, which is involved in amino acid metabolism; and Tsl1, which mediates carbohydrate storage. Finally, we identified the TORC1 substrates Sch9 and Atg1 as candidate kinases that phosphorylate Amd1 and Hom3, respectively.

An automated pipeline for high-throughput label-free quantitative proteomics.

We present a computational pipeline for the quantification of peptides and proteins in label-free LC-MS/MS data sets. The pipeline is composed of tools from the OpenMS software framework and is applicable to the processing of large experiments (50+ samples). We describe several enhancements that we have introduced to OpenMS to realize the implementation of this pipeline. They include new algorithms for centroiding of raw data, for feature detection, for the alignment of multiple related measurements, and a new tool for the calculation of peptide and protein abundances. Where possible, we compare the performance of the new algorithms to that of their established counterparts in OpenMS. We validate the pipeline on the basis of two small data sets that provide ground truths for the quantification. There, we also compare our results to those of MaxQuant and Progenesis LC-MS, two popular alternatives for the analysis of label-free data. We then show how our software can be applied to a large heterogeneous data set of 58 LC-MS/MS runs.

Streptococcus pyogenes in human plasma: adaptive mechanisms analyzed by mass spectrometry-based proteomics.

Streptococcus pyogenes is a major bacterial pathogen and a potent inducer of inflammation causing plasma leakage at the site of infection. A combination of label-free quantitative mass spectrometry-based proteomics strategies were used to measure how the intracellular proteome homeostasis of S. pyogenes is influenced by the presence of human plasma, identifying and quantifying 842 proteins. In plasma the bacterium modifies its production of 213 proteins, and the most pronounced change was the complete down-regulation of proteins required for fatty acid biosynthesis. Fatty acids are transported by albumin (HSA) in plasma. S. pyogenes expresses HSA-binding surface proteins, and HSA carrying fatty acids reduced the amount of fatty acid biosynthesis proteins to the same extent as plasma. The results clarify the function of HSA-binding proteins in S. pyogenes and underline the power of the quantitative mass spectrometry strategy used here to investigate bacterial adaptation to a given environment.

Only slight impact of predicted replicative capacity for therapy response prediction.

BACKGROUND: Replication capacity (RC) of specific HIV isolates is occasionally blamed for unexpected treatment responses. However, the role of viral RC in response to antiretroviral therapy is not yet fully understood. MATERIALS AND METHODS: We developed a method for predicting RC from genotype using support vector machines (SVMs) trained on about 300 genotype-RC pairs. Next, we studied the impact of predicted viral RC (pRC) on the change of viral load (VL) and CD4(+) T-cell count (CD4) during the course of therapy on about 3,000 treatment change episodes (TCEs) extracted from the EuResist integrated database. Specifically, linear regression models using either treatment activity scores (TAS), the drug combination, or pRC or any combination of these covariates were trained to predict change in VL and CD4, respectively. RESULTS: The SVM models achieved a Spearman correlation (rho) of 0.54 between measured RC and pRC. The prediction of change in VL (CD4) was best at 180 (360) days, reaching a correlation of rho = 0.45 (rho = 0.27). In general, pRC was inversely correlated to drug resistance at treatment start (on average rho = -0.38). Inclusion of pRC in the linear regression models significantly improved prediction of virological response to treatment based either on the drug combination or on the TAS (t-test; p-values range from 0.0247 to 4 10(-6)) but not for the model using both TAS and drug combination. For predicting the change in CD4 the improvement derived from inclusion of pRC was not significant. CONCLUSION: Viral RC could be predicted from genotype with moderate accuracy and could slightly improve prediction of virological treatment response. However, the observed improvement could simply be a consequence of the significant correlation between pRC and drug resistance.