May Institute Essentials

Modern DIA experiments can quantify thousands of proteins, but turning that output into conclusions you can trust is where many analyses go wrong. In this session we’ll walk through a complete workflow, from the reports of spectral processing tools like DIA-NN and Spectronaut to differential abundance results. We’ll cover the decisions that most affect your conclusions (normalization, missing values, feature selection, and protein summarization) and show how to choose the right statistical model for your design, including time courses, paired samples, and experimental covariates. We’ll use MSstats throughout, but the principles apply to any proteomics analysis tool. All hands-on work uses point-and-click web interfaces, so no programming experience is needed. You’ll leave able to take your own data from raw output to defensible results.

Post-translational modification (PTM) and chemoproteomics experiments provide important insights into protein regulation and drug‐protein interactions, but introduce unique challenges for statistical analysis. PTM experiments often have limited measurements at individual modification sites and require distinguishing changes in PTM abundance from changes in overall protein abundance, while chemoproteomics experiments measure protein responses across multiple drug concentrations and may not follow a standard dose-response curve shape. In the first part of the session, we introduce MSstatsPTM, a tool for detecting differential PTM abundance while accounting for changes in overall protein abundance. The second part focuses on statistical modeling for chemoproteomics experiments with MSstatsResponse, which applies flexible dose-response modeling to protein‐ or PTM‐level data to reliably detect drug‐protein interactions, estimate IC50 values, and visualize response curves. Through hands-on examples in our RShiny GUIs, participants will gain practical skills for analyzing both experiment types.

Interpreting proteomics data using biological networks representing cellular mechanisms is a powerful approach to gaining actionable insights. In this course, we first introduce the INDRA system, developed by the Gyori Lab, which automatically assembles mechanisms into networks from both pathway databases and text-mined biomedical literature. We then dive deeper into the INDRA Database, the INDRA Network Search tool, and the INDRA Biomedical Discovery Engine which each facilitate gaining insights from data and generating hypotheses based on large-scale networks. Finally, we introduce MSstatsBioNet, which queries INDRA to construct an experiment-specific biological subnetwork from statistical results of MS proteomics experiments, facilitating biological interpretation. The course will be conducted using web-based UIs and RShiny GUIs.

In this session we will cover the basics of Cardinal, beginning with Cardinal's data structures and the .imzML format as well as the basics of working with larger-than-memory datasets. We will then cover preprocessing of MSI data, including recalibration, peak picking, peak alignment, and normalization as well as common operations like subsetting and summarizing pixels and features. We will end with visualization of spectra and single ion images.

This session will cover analysis of preprocessed MSI data. First we will cover region of interest (ROI) discovery using Cardinal's univariate and multivariate segmentation methods. We will then conclude with a case study of differential abundance between ROIs and experimental conditions.