A fundamental task in precision oncology is to perform mutation profiling of tumors, which traditionally requires invasive tumor biopsy. Liquid biopsy, which requires a simple blood draw, provides a surrogate option for tumor biopsy. Liquid biopsy is noninvasive, repeatable, and especially useful when a tumor’s location makes tissue biopsy unfeasible.

However, the mutation profiling of cell-free DNA (cfDNA) is challenging, as conventional mutation callers are primarily designed for the mutation calling of genomic DNA. Unlike tumor biopsy, cfDNA is a mixture of DNA derived from tumors and that from normal cells, where tumor-derived cfDNA accounts for a minor fraction. As a result, mutations from tumors have weak signals in blood samples and are therefore hard to identify using conventional methods.

To address this challenge, in our recent work published in Nature Communications [1], we present cfSNV, a cfDNA somatic single nucleotide variant (SNV) caller with five innovative techniques to overcome and exploit the unique properties of cfDNA. Specifically, (1) cfDNA is a mixture of tumor and normal DNAs, so we developed a joint-genotype statistical model to increase the sensitivity of the mutation calling; (2) cfDNA is heterogenous, as it contains mutations from all tumor clones and subclones, so we designed a hierarchical and iterative procedure to comprehensively call mutations; (3) cfDNA fragments are short and their size peaks at approximately 160 base pairs, so we utilized this property to add an error suppression step to improve the accuracy of mutation calls; (4) the fragmentation of cfDNA is nonrandom, so we designed a new filtering procedure to prevent the incorrect removal of true mutations; and (5) tumor cfDNA often occupies a very minor fraction of all cfDNA, so we employed a machine learning approach to ensure the quality of our mutation calls even at a low variant allele frequency (VAF).

In this study, we showed that these five techniques resulted in cfSNV outperforming existing methods in terms of sensitivity and specificity and in reliably detecting mutations in cfDNA, even with medium-coverage whole-exome sequencing (WES) data. Compared to conventional callers, for all mutations, cfSNV achieves around 30% increase in sensitivity while remaining comparable in precision or improving precision;  for low VAF mutations, cfSNV achieves around 40%  improvement in sensitivity without sacrificing precision.

The applicability of cfSNV to medium-coverage (e.g., 200x) WES data enables cfDNA to be used in a wide variety of clinical applications, and we present one example application: a novel and effective immunotherapy response measure that we named truncal-bTMB. This measure exploits the ability of cfDNA to provide comprehensive coverage of the mutation landscape. We demonstrated that this new marker allowed effective immunotherapy prognostication by simultaneously capturing both the tumor mutation burden and clonal structure information.

In summary, the mutation profiling of cancer patients from liquid biopsy offers a unique opportunity to noninvasively characterize biomarkers and guide cancer treatment. cfSNV will expand the usage of mutation profiling from liquid biopsy and facilitate noninvasive personalized medicine.

[1] Li, Shuo, et al. "Sensitive detection of tumor mutations from blood and its application to immunotherapy prognosis." Nature communications 12.1 (2021): 1-14.