Abstract: Methods for determining a DNA mismatch repair deficiency (MMRD) status for a sample from an individual are described. The methods may comprise, for example, receiving sequence read data for a plurality of sequence reads derived from a sample from an individual; extracting two or more genomic features of the sample based on the sequence read data; determining a MMRD probability score for the sample based on the two or more extracted genomic features and a machine learning model configured to process genomic feature input data and output MMRD probability scores; and comparing the MMRD probability score to one or more predetermined thresholds to determine a MMRD status for the sample from the individual.

Abstract: Disclosed are compositions and methods related to the use of plurality of reverse transcriptase primers, unique molecular identifiers (UMIs), and/or spiky primers with unique junction identifiers to improve the sequencing and amplifications methods. In some embodiments, the disclosed methods can identify sequencing errors and PCR-jumping errors.