Vardetector: a pure Python package to detect DNA called mutations in aligned RNA reads

We develop a freely-available Python package Vardetector (https://github.com/julijselb/vardetector/tree/main/vardetector) used for detecting DNA called mutations in aligned RNA reads. We benchmark it by comparing it to industry standard variant caller (GATK HaplotypeCaller; r = 0.88896/0.88859 (supporting-reads/all-reads)) and demonstrate the functionality by comparing two RNA-seq library preparation protocols for formalin fixed paraffin embedded (FFPE) tumor samples. One protocol relies on exome-capture and the other on ribosome-depletion (ribodepletion) chemistry. We call somatic mutations from DNA of tumor/normal samples of two individuals with non-small cell lung cancer and test the difference between the two protocols by quantifying all RNA reads (all-reads) and somatic mutation supporting RNA reads (supporting-reads) over the positions of the DNA-called mutations. We show that the ribodepletion protocol produces significantly higher number of all (p < 0.001) and of supporting (p < 0.001) reads over the mutations of interest. Moreover, the ribodepletion protocol produces significantly (p < 0.001) wider breath of somatic mutation position coverage. The Vardetector software package and our results display a meaningful potential of the approach to improve neoantigen prioritisation pipelines.