Search

We present a deep-learning-based platform, MIND-S, for protein post-translational modification (PTM) predictions. MIND-S employs a multi-head attention and graph neural network and assembles a 15-fold ensemble model in a multi-label strategy to enable simultaneous prediction of multiple PTMs with high performance and computation efficiency. MIND-S also features an interpretation module, which provides the relevance of each amino acid for making the predictions and is validated with known motifs. The interpretation module also captures PTM patterns without any supervision. Furthermore, MIND-S enables examination of mutation effects on PTMs. We document a workflow, its applications to 26 types of PTMs of two datasets consisting of ∼50,000 proteins, and an example of MIND-S identifying a PTM-interrupting SNP with validation from biological data. We also include use case analyses of targeted proteins. Taken together, we have demonstrated that MIND-S is accurate, interpretable, and efficient to elucidate PTM-relevant biological processes in health and diseases.

Macrophage autophagy is a highly anti-atherogenic process that promotes the catabolism of cytosolic lipid droplets (LDs) to maintain cellular lipid homeostasis. Selective autophagy relies on tags such as ubiquitin and a set of selectivity factors including selective autophagy receptors (SARs) to label specific cargo for degradation. Originally described in yeast cells, “lipophagy” refers to the degradation of LDs by autophagy. Yet, how LDs are targeted for autophagy is poorly defined. Here, we employed mass spectrometry to identify lipophagy factors within the macrophage foam cell LD proteome. In addition to structural proteins (e.g., PLIN2), metabolic enzymes (e.g., ACSL) and neutral lipases (e.g., PNPLA2), we found the association of proteins related to the ubiquitination machinery (e.g., AUP1) and autophagy (e.g., HMGB, YWHA/14-3-3 proteins). The functional role of candidate lipophagy factors (a total of 91) was tested using a custom siRNA array combined with high-content cholesterol efflux assays. We observed that knocking down several of these genes, including Hmgb1, Hmgb2, Hspa5, and Scarb2, significantly reduced cholesterol efflux, and SARs SQSTM1/p62, NBR1 and OPTN localized to LDs, suggesting a role for these in lipophagy. Using yeast lipophagy assays, we established a genetic requirement for several candidate lipophagy factors in lipophagy, including HSPA5, UBE2G2 and AUP1. Our study is the first to systematically identify several LD-associated proteins of the lipophagy machinery, a finding with important biological and therapeutic implications. Targeting these to selectively enhance lipophagy to promote cholesterol efflux in foam cells may represent a novel strategy to treat atherosclerosis. Abbreviations: ADGRL3: adhesion G protein-coupled receptor L3; agLDL: aggregated low density lipoprotein; AMPK: AMP-activated protein kinase; APOA1: apolipoprotein A1; ATG: autophagy related; AUP1: AUP1 lipid droplet regulating VLDL assembly factor; BMDM: bone-marrow derived macrophages; BNIP3L: BCL2/adenovirus E1B interacting protein 3-like; BSA: bovine serum albumin; CALCOCO2: calcium binding and coiled-coil domain 2; CIRBP: cold inducible RNA binding protein; COLGALT1: collagen beta(1-O)galactosyltransferase 1; CORO1A: coronin 1A; DMA: deletion mutant array; Faa4: long chain fatty acyl-CoA synthetase; FBS: fetal bovine serum; FUS: fused in sarcoma; HMGB1: high mobility group box 1; HMGB2: high mobility group box 2: HSP90AA1: heat shock protein 90: alpha (cytosolic): class A member 1; HSPA5: heat shock protein family A (Hsp70) member 5; HSPA8: heat shock protein 8; HSPB1: heat shock protein 1; HSPH1: heat shock 105kDa/110kDa protein 1; LDAH: lipid droplet associated hydrolase; LIPA: lysosomal acid lipase A; LIR: LC3-interacting region; MACROH2A1: macroH2A.1 histone; MAP1LC3: microtubule-associated protein 1 light chain 3; MCOLN1: mucolipin 1; NBR1: NBR1, autophagy cargo receptor; NPC2: NPC intracellular cholesterol transporter 2; OPTN: optineurin; P/S: penicillin-streptomycin; PLIN2: perilipin 2; PLIN3: perilipin 3; PNPLA2: patatin like phospholipase domain containing 2; RAB: RAB, member RAS oncogene family; RBBP7, retinoblastoma binding protein 7, chromatin remodeling factor; SAR: selective autophagy receptor; SCARB2: scavenger receptor class B, member 2; SGA: synthetic genetic array; SQSTM1: sequestosome 1; TAX1BP1: Tax1 (human T cell leukemia virus type I) binding protein 1; TFEB: transcription factor EB; TOLLIP: toll interacting protein; UBE2G2: ubiquitin conjugating enzyme E2 G2; UVRAG: UV radiation resistance associated gene; VDAC2: voltage dependent anion channel 2; VIM: vimentin

Objective During cardiovascular disease progression, molecular systems of myocardium (e.g., a proteome) undergo diverse and distinct changes. Dynamic, temporally-regulated alterations of individual molecules underlie the collective response of the heart to pathological drivers and the ultimate development of pathogenesis. Advances in high-throughput omics technologies have enabled cost-effective, temporal profiling of targeted systems in animal models of human diseases. However, computational analysis of temporal patterns from omics data remains challenging. In particular, bioinformatic pipelines involving unsupervised statistical approaches to support cardiovascular investigations are lacking, which hinders one's ability to extract biomedical insights from these complex datasets. Approach and results We developed a non-parametric data analysis platform to resolve computational challenges unique to temporal omics datasets. Our platform consists of three modules. Module I preprocesses the temporal data using either cubic splines or principal component analysis (PCA), and it simultaneously accomplishes the tasks on missing data imputation and denoising. Module II performs an unsupervised classification by K-means or hierarchical clustering. Module III evaluates and identifies biological entities (e.g., molecular events) that exhibit strong associations to specific temporal patterns. The jackstraw method for cluster membership has been applied to estimate p-values and posterior inclusion probabilities (PIPs), both of which guided feature selection. To demonstrate the utility of the analysis platform, we employed a temporal proteomics dataset that captured the proteome-wide dynamics of oxidative stress induced post-translational modifications (O-PTMs) in mouse hearts undergoing isoproterenol (ISO)-induced hypertrophy. Conclusion We have created a platform, CV.Signature.TCP, to identify distinct temporal clusters in omics datasets. We presented a cardiovascular use case to demonstrate its utility in unveiling biological insights underlying O-PTM regulations in cardiac remodeling. This platform is implemented in an open source R package ( https://github.com/UCLA-BD2K/CV.Signature.TCP ).

Mass spectrometry-based proteomics technologies are the prime methods for the high-throughput identification of proteins in complex biological samples. Nevertheless, there are still technical limitations that hinder the ability of mass spectrometry to identify low abundance proteins in complex samples. Characterizing such proteins is essential to provide a comprehensive understanding of the biological processes taking place in cells and tissues. Still today, most mass spectrometry-based proteomics approaches use a data-dependent acquisition strategy, which favors the collection of mass spectra from proteins of higher abundance. Since the computational identification of proteins from proteomics data is typically performed after mass spectrometry analysis, large numbers of mass spectra are typically redundantly acquired from the same abundant proteins, and little to no mass spectra are acquired for proteins of lower abundance. We therefore propose a novel supervised learning algorithm that identifies proteins in real-time as mass spectrometry data are acquired and prevents further data collection from confidently identified proteins to ultimately free mass spectrometry resources to improve the identification sensitivity of low abundance proteins. We use real-time simulations of a previously performed mass spectrometry analysis of a HEK293 cell lysate to show that our approach can identify 92.1% of the proteins detected in the experiment using 66.2% of the MS2 spectra. We also demonstrate that our approach outperforms a previously proposed method, is sufficiently fast for real-time mass spectrometry analysis, and is flexible. Finally, MealTime-MS’ efficient usage of mass spectrometry resources will provide a more comprehensive characterization of proteomes in complex samples.

When coupled to mass spectrometry (MS), energetics-based protein separation (EBPS) techniques, such as thermal shift assay, have shown great potential to identify the targets of a drug on a proteome scale. Nevertheless, the computational analyses assessing the confidence of drug target predictions made by these methods have remained rudimentary and significantly differ depending on the protocol used to produce the data. To identify drug targets in datasets produced using different EBPS-MS techniques, we have developed a novel flexible computational approach named TargetSeeker-MS. We showed that TargetSeeker-MS reproducibly identifies known and novel drug targets inC. elegansand HEK293 samples that were treated with the fungicide benomyl and processed using two different EBPS techniques. We also validated a novel benomyl target in vitro. TargetSeeker-MS, which is available online, allows for the confident identification of targets of a drug on a proteome scale, thereby facilitating the evaluation of its clinical viability.

Kinases are a major clinical target for human diseases. Identifying the proteins that interact with kinases in vivo will provide information on unreported substrates and will potentially lead to more specific methods for therapeutic kinase regulation. Here, endogenous immunoprecipitations of evolutionally distinct kinases (i.e., Akt, ERK2, and CAMK2) from rodent hippocampi were analyzed by mass spectrometry to generate three highly confident kinase protein-protein interaction networks. Proteins of similar function were identified in the networks, suggesting a universal model for kinase signaling complexes. Protein interactions were observed between kinases with reported symbiotic relationships. The kinase networks were significantly enriched in genes associated with specific neurodevelopmental disorders providing novel structural connections between these disease-associated genes. To demonstrate a functional relationship between the kinases and the network, pharmacological manipulation of Akt in hippocampal slices was shown to regulate the activity of potassium/sodium hyperpolarization-activated cyclic nucleotide-gated channel(HCN1), which was identified in the Akt network. Overall, the kinase protein-protein interaction networks provide molecular insight of the spatial complexity of in vivo kinase signal transduction which is required to achieve the therapeutic potential of kinase manipulation in the brain.

Background Gene fusions often occur in cancer cells and in some cases are the main driver of oncogenesis. Correct identification of oncogenic gene fusions thus has implications for targeted cancer therapy. Recognition of this potential has led to the development of a myriad of sequencing-based fusion detection tools. However, given the same input, many of these detectors will find different fusion points or claim different sets of supporting data. Furthermore, the rate at which these tools falsely detect fusion events in data varies greatly. This discrepancy between tools underscores the fact that computation algorithms still cannot perfectly evaluate evidence; especially when provided with small amounts of supporting data as is typical in fusion detection. We assert that when evidence is provided in an easily digestible form, humans are more proficient in identifying true positives from false positives. Results We have developed a web tool that, given the genomic coordinates of a candidate fusion breakpoint, will extract fusion and non-fusion reads adjacent to the fusion point from partner transcripts, and color code reads by transcript origin and read orientation for ease of intuitive inspection by the user. Fusion partner transcript read alignments are performed using a novel variant of the Smith-Waterman algorithm. Conclusions Combined with dynamic filtering parameters, the visualization provided by our tool introduces a powerful new investigative step that allows researchers to comprehensively evaluate fusion evidence. Additionally, this allows quick identification of false positives that may deceive most fusion detectors, thus eliminating unnecessary gene fusion validation. We apply our visualization tool to publicly available datasets and provide examples of true as well as false positives reported by open source fusion detection tools.

QuaCRS ( Quality Control for RNA- Seq) is an integrated, simplified quality control (QC) system for RNA-seq data that allows easy execution of several open-source QC tools, aggregation of their output, and the ability to quickly identify quality issues by performing meta-analyses on QC metrics across large numbers of samples in different studies. It comprises two main sections. First is the QC Pack wrapper, which executes three QC tools: FastQC, RNA-SeQC, and selected functions from RSeQC. Combining these three tools into one wrapper provides increased ease of use and provides a much more complete view of sample data quality than any individual tool. Second is the QC database, which displays the resulting metrics in a user-friendly web interface. It was designed to allow users with less computational experience to easily generate and view QC information for their data, to investigate individual samples and aggregate reports of sample groups, and to sort and search samples based on quality. The structure of the QuaCRS database is designed to enable expansion with additional tools and metrics in the future. The source code for not-for-profit use and a fully functional sample user interface with mock data are available at http://bioserv.mps.ohio-state.edu/QuaCRS/ .