A machine learning approach to optimizing cell-free DNA sequencing panels: with an application to prostate cancer

Abstract

Background: Cell-free DNA's (cfDNA) use as a biomarker in cancer is challenging due to genetic heterogeneity of malignancies and rarity of tumor-derived molecules. Here we describe and demonstrate a novel machine-learning guided panel design strategy for improving the detection of tumor variants in cfDNA. Using this approach, we first generated a model to classify and score candidate variants for inclusion on a prostate cancer targeted sequencing panel. We then used this panel to screen tumor variants from prostate cancer patients with localized disease in both in silico and hybrid capture settings. Methods: Whole Genome Sequence (WGS) data from 550 prostate tumors was analyzed to build a targeted sequencing panel of single point and small (< 200 bp) indel mutations, which was subsequently screened in silico against prostate tumor sequences from 5 patients to assess performance against commonly used alternative panel designs. The panel's ability to detect tumor-derived cfDNA variants was then assessed using prospectively collected cfDNA and tumor foci from a test set 18 prostate cancer patients with localized disease undergoing radical proctectomy. Results: The panel generated from this approach identified as top candidates mutations in known driver genes (e.g. HRAS) and prostate cancer related transcription factor binding sites (e.g. MYC, AR). It outperformed two commonly used designs in detecting somatic mutations found in the cfDNA of 5 prostate cancer patients when analyzed in an in silico setting. Additionally, hybrid capture and 2500X sequencing of cfDNA molecules using the panel resulted in detection of tumor variants in all 18 patients of a test set, where 15 of the 18 patients had detected variants found in multiple foci. Conclusion: Machine learning-prioritized targeted sequencing panels may prove useful for broad and sensitive variant detection in the cfDNA of heterogeneous diseases. This strategy has implications for disease detection and monitoring when applied to the cfDNA isolated from prostate cancer patients.

Integrating chromatin accessibility states in the design of targeted sequencing panels for liquid biopsy

“…Typically, targeted sequencing panels for liquid biopsy are designed to cover recurrent somatic mutations 4 , 8 , 9 or DMR regions 18 , 19 specific to one or more cancer types of interest. It has previously been established however, that even when ctDNA is detected with sensitive methods, only a subset of somatic mutations that are found in tumors are also called in the plasma 5 , 14 , 41 . Whether or not a tumor-associated variant is faithfully reflected in the ctDNA variant pool depends on multiple known and perhaps some unknown factors 14 , 19 .…”

“…It has previously been established however, that even when ctDNA is detected with sensitive methods, only a subset of somatic mutations that are found in tumors are also called in the plasma 5 , 14 , 41 . Whether or not a tumor-associated variant is faithfully reflected in the ctDNA variant pool depends on multiple known and perhaps some unknown factors 14 , 19 .…”

“…Here we ranked cancer specific mutations 43 – 45 and tDMRs from 2 , 42 , (Supplementary Data 5 ). ( b ) Matched tumor and cfDNA targeted sequencing data from 23 prostate cancer patients (obtained from 14 ) were processed to variant calls (“ Methods ”). For each patient, variants detected in both tumor sample and cfDNA were selected and then divided to two groups based on their corresponding RelAccS score (RelAccS > 0 (pink), RelAccS < 0 (green)).…”

“…To validate the utility of using RelAccS for marker prioritization, we evaluated cell-free DNA of cancer patients with matched tumor in prostate cancer 14 . We hypothesized that since DNA fragments from regions with low accessibility in tumor but high accessibility in blood (RelAccS < 0) should be better represented in cfDNA than those with the opposite pattern (RelAccS > 0), allele frequencies in cfDNA should better reflect those in matched tumors for the former scenario.…”

“…However, this limits the utility of liquid biopsy to diseases associated with somatic genetic aberrations, and even among those, only to subtypes with multiple known recurrent mutations or known driver genes. To improve the sensitivity of mutation detection, several diverse methods have been proposed, including barcoding fragments with unique molecular indices (UMI) followed by error suppression 12 , using the phasing of multiple somatic mutations in individual DNA fragments 13 , and machine-learning based scoring to prioritize driver mutations for inclusion in panel designs 14 .…”