Post-translational modifications (PTMs) play a critical role in many aspects of cell biology including cell growth, differentiation, and survival. Accurate identifications of PTM peptides and their PTM locations are very important in proteomics. The retention time information of PTM peptides can be used to improve the accuracy of PTM identification and experimental proteomic strategies. Therefore, understanding PTM effects on retention times will be beneficial for future research.Here, we present VA-PRT (Visualization tool for Analyzing Post-translational modification Retention Times) that systematically investigates the PTM effects on retention times using visualization techniques and advanced statistical methods. The usefulness of VA-PRT is demonstrated using a large synthetic proteomic and phosphoproteomic dataset. VA-PRT is implemented in R and will be freely available for public use.

Existing Traumatic Brain Injury (TBI) mortality prediction models have not historically accounted for the low mortality rate of pediatric patients, causing class-imbalance problems and undermining classifier prediction performance. We minimize class-imbalance effects by separately implementing case-weighting and subsampling for eight supervised machine learning (ML) techniques, trained and evaluated on pediatric patient National Trauma Data Bank (NTDB) data. The training set consists of pediatric TBI patient records between 2010 and 2014, where records from 2015 are used as the test set. Unweighted models, trained on data modified by Synthetic Minority Over-sampling Technique (SMOTE) and Random Over-sampling Examples (ROSE), included Stacked Autoencoder Deep Neural Network (SAE-DNN), K-Nearest Neighbors (KNN), Naive Bayes Classifier (NB), and Support Vector Machine (SVM). Case-weighted models, trained on standardized data, included Artificial Neural Network (ANN), Extreme Gradient Boosting (XGB), C5.0 Decision Tree (C5.0), and Logistic Regression (LR). Preliminarily, XGB and C5.0 were best able to distinguish between classes. Models trained on SMOTE data maximized metrics accounting for class-imbalance and models trained on ROSE data tended to generally underperform.

Abstract Contact: [email protected]

MOTIVATION: Mass spectrometry (MS)-based high-throughput quantitative proteomics shows great potential in large-scale clinical biomarker studies, identifying and quantifying thousands of proteins in biological samples. However, there are unique challenges in analyzing the quantitative proteomics data. One issue is that the quantification of a given peptide is often missing in a subset of the experiments, especially for less abundant peptides. Another issue is that different MS experiments of the same study have significantly varying numbers of peptides quantified, which can result in more missing peptide abundances in an experiment that has a smaller total number of quantified peptides. To detect as many biomarker proteins as possible, it is necessary to develop bioinformatics methods that appropriately handle these challenges. RESULTS: We propose a Significance Analysis for Large-scale Proteomics Studies (SALPS) that handles missing peptide intensity values caused by the two mechanisms mentioned above. Our model has a robust performance in both simulated data and proteomics data from a large clinical study. Because varying patients' sample qualities and deviating instrument performances are not avoidable for clinical studies performed over the course of several years, we believe that our approach will be useful to analyze large-scale clinical proteomics data. AVAILABILITY AND IMPLEMENTATION: R codes for SALPS are available at http://www.stanford.edu/%7eclairesr/software.html.