QuantUMS: Machine learning-based quantification method improves accuracy and precision in DIA proteomics (nature.com)

• QuantUMS dynamically optimizes quantification hyperparameters to minimize errors and reduce ratio compression in DIA proteomics.
• Integrates all MS1 and MS/MS signals, providing uncertainty estimates for each quantified precursor and protein.
• Outperforms legacy DIA-NN quantification on multiple benchmarks, increasing precision and sensitivity in differential expression analysis.

"QuantUMS, a machine learning algorithm integrated into DIA-NN, dynamically tunes quantification to minimize errors in data-independent acquisition (DIA) proteomics. It integrates all MS1 and MS/MS signals, reducing ratio compression bias and providing uncertainty estimates for each quantity. Benchmarks on mixed-species, human fibroblast, and chronic lymphocytic leukemia datasets show improved accuracy, precision, and differential expression sensitivity compared to legacy DIA-NN quantification. The method is available as a standalone tool and within DIA-NN."