Error-Corrected Deep Targeted Sequencing of Circulating Cell-Free DNA from Colorectal Cancer Patients for Sensitive Detection of Circulating Tumor DNA.

Circulating tumor DNA (ctDNA) is a promising biomarker, reflecting the presence of tumor cells. Sequencing-based detection of ctDNA at low tumor fractions is challenging due to the crude error rate of sequencing. To mitigate this challenge, we developed ultra-deep mutation-integrated sequencing (UMIseq), a fixed-panel deep targeted sequencing approach, which is universally applicable to all colorectal cancer (CRC) patients. UMIseq features UMI-mediated error correction, the exclusion of mutations related to clonal hematopoiesis, a panel of normal samples for error modeling, and signal integration from single-nucleotide variations, insertions, deletions, and phased mutations. UMIseq was trained and independently validated on pre-operative (pre-OP) plasma from CRC patients (n = 364) and healthy individuals (n = 61). UMIseq displayed an area under the curve surpassing 0.95 for allele frequencies (AFs) down to 0.05%. In the training cohort, the pre-OP detection rate reached 80% at 95% specificity, while it was 70% in the validation cohort. UMIseq enabled the detection of AFs down to 0.004%. To assess the potential for detection of residual disease, 26 post-operative plasma samples from stage III CRC patients were analyzed. From this we found that the detection of ctDNA was associated with recurrence. In conclusion, UMIseq demonstrated robust performance with high sensitivity and specificity, enabling the detection of ctDNA at low allele frequencies.

Beyond basics: Key mutation selection features for successful tumor-informed ctDNA detection.

Tumor-informed mutation-based approaches are frequently used for detection of circulating tumor DNA (ctDNA). Not all mutations make equally effective ctDNA markers. The objective was to explore if prioritizing mutations using mutational features-such as cancer cell fraction (CCF), multiplicity, and error rate-would improve the success rate of tumor-informed ctDNA analysis. Additionally, we aimed to develop a practical and easily implementable analysis pipeline for identifying and prioritizing candidate mutations from whole-exome sequencing (WES) data. We analyzed WES and ctDNA data from three tumor-informed ctDNA studies, one on bladder cancer (Cohort A) and two on colorectal cancer (Cohorts I and N). The studies included 390 patients. For each patient, a unique set of mutations (median mutations/patient: 6, interquartile 13, range: 1-46, total n = 4023) were used as markers of ctDNA. The tool PureCN was used to assess the CCF and multiplicity of each mutation. High-CCF mutations were detected more frequently than low-CCF mutations (Cohort A: odds ratio [OR] 20.6, 95% confidence interval [CI] 5.72-173, p = 1.73e-12; Cohort I: OR 2.24, 95% CI 1.44-3.52, p = 1.66e-04; and Cohort N: OR 1.78, 95% CI 1.14-2.79, p = 7.86e-03). The detection-likelihood was additionally improved by selecting mutations with multiplicity of two or above (Cohort A: OR 1.55, 95% CI 1. 14-2.11, p = 3.85e-03; Cohort I: OR 1.78, 95% CI 1.23-2.56, p = 1.34e-03; and Cohort N: OR 1.94, 95% CI 1.63-2.31, p = 2.83e-14). Furthermore, selecting the mutations for which the ctDNA detection method had the lowest error rates, additionally improved the detection-likelihood, particularly evident when plasma cell-free DNA tumor fractions were below 0.1% (p = 2.1e-07). Selecting mutational markers with high CCF, high multiplicity, and low error rate significantly improve ctDNA detection likelihood. We provide free access to the analysis pipeline enabling others to perform qualified prioritization of mutations for tumor-informed ctDNA analysis.

DREAMS: deep read-level error model for sequencing data applied to low-frequency variant calling and circulating tumor DNA detection.

Circulating tumor DNA detection using next-generation sequencing (NGS) data of plasma DNA is promising for cancer identification and characterization. However, the tumor signal in the blood is often low and difficult to distinguish from errors. We present DREAMS (Deep Read-level Modelling of Sequencing-errors) for estimating error rates of individual read positions. Using DREAMS, we develop statistical methods for variant calling (DREAMS-vc) and cancer detection (DREAMS-cc). For evaluation, we generate deep targeted NGS data of matching tumor and plasma DNA from 85 colorectal cancer patients. The DREAMS approach performs better than state-of-the-art methods for variant calling and cancer detection.

Impact of Whole Genome Doubling on Detection of Circulating Tumor DNA in Colorectal Cancer.

OBJECTIVE: Circulating tumor DNA (ctDNA) is a candidate biomarker of cancer with practice-changing potential in the detection of both early and residual disease. Disease stage and tumor size affect the probability of ctDNA detection, whereas little is known about the influence of other tumor characteristics on ctDNA detection. This study investigates the impact of tumor cell whole-genome doubling (WGD) on the detection of ctDNA in plasma collected preoperatively from newly diagnosed colorectal cancer (CRC) patients. METHODS: WGD was estimated from copy numbers derived from whole-exome sequencing (WES) data of matched tumor and normal DNA from 833 Danish CRC patients. To explore if tumor WGD status impacts ctDNA detection, we applied tumor-informed ctDNA analysis to preoperative plasma samples from all patients. RESULTS: Patients with WGD+ tumors had 53% increased odds of being ctDNA positive (OR = 1.53, 95%CI: 1.12-2.09). After stratification for UICC stage, the association persisted for Stage I (OR = 2.44, 95%CI: 1.22-5.03) and Stage II (OR = 1.76, 95%CI: 1.11-2.81) but not for Stage III (OR = 0.83, 95%CI: 0.44-1.53) patients. CONCLUSION: The presence of WGD significantly increases the probability of detecting ctDNA, particularly for early-stage disease. In patients with more advanced disease, the benefit of WGD on ctDNA detection is less pronounced, consistent with increased DNA shedding from these tumors, making ctDNA detection less dependent on the amount of ctDNA released per tumor cell.

Error Characterization and Statistical Modeling Improves Circulating Tumor DNA Detection by Droplet Digital PCR.

BACKGROUND: Droplet digital PCR (ddPCR) is a widely used and sensitive application for circulating tumor DNA (ctDNA) detection. As ctDNA is often found in low abundance, methods to separate low-signal readouts from noise are necessary. We aimed to characterize the ddPCR-generated noise and, informed by this, create a sensitive and specific ctDNA caller. METHODS: We built 2 novel complimentary ctDNA calling methods: dynamic limit of blank and concentration and assay-specific tumor load estimator (CASTLE). Both methods are informed by empirically established assay-specific noise profiles. Here, we characterized noise for 70 mutation-detecting ddPCR assays by applying each assay to 95 nonmutated samples. Using these profiles, the performance of the 2 new methods was assessed in a total of 9447 negative/positive reference samples and in 1311 real-life plasma samples from colorectal cancer patients. Lastly, performances were compared to 7 literature-established calling methods. RESULTS: For many assays, noise increased proportionally with the DNA input amount. Assays targeting transition base changes were more error-prone than transversion-targeting assays. Both our calling methods successfully accounted for the additional noise in transition assays and showed consistently high performance regardless of DNA input amount. Calling methods that were not noise-informed performed less well than noise-informed methods. CASTLE was the only calling method providing a statistical estimate of the noise-corrected mutation level and call certainty. CONCLUSIONS: Accurate error modeling is necessary for sensitive and specific ctDNA detection by ddPCR. Accounting for DNA input amounts ensures specific detection regardless of the sample-specific DNA concentration. Our results demonstrate CASTLE as a powerful tool for ctDNA calling using ddPCR.

Tumour-agnostic circulating tumour DNA analysis for improved recurrence surveillance after resection of colorectal liver metastases: A prospective cohort study.

PURPOSE: Nearly 50% of patients recur within two years after curatively intended resection of colorectal cancer liver metastasis (CRLM). The optimal surveillance strategy is unknown due to the lack of evidence. Here, we explored the potential for improving postoperative CRLM surveillance by performing serial circulating tumour DNA (ctDNA) assessments parallel to standard-of-care surveillance. EXPERIMENTAL DESIGN: 499 prospectively collected serial plasma samples from 96 patients undergoing CRLM resection were analysed using the tumour-agnostic methylation multiplex droplet-digital PCR test 'TriMeth'. RESULTS: Patients with ctDNA postoperatively or post adjuvant chemotherapy experienced a significant lower recurrence-free survival than patients without ctDNA (hazard ratio (HR) 4.5; P < 0.0001 and HR 8.4, P < 0.0001). ctDNA status was a stronger predictor of recurrence than standard clinical risk factors and carcinoembryonic antigen. Serial TriMeth analysis detected ctDNA before radiological recurrence in 55.6% of ctDNA-positive patients, with up to 10.6 months lead-time (median 3.1 months). During surveillance, 24% of patients had inconclusive CT scans, which was associated with a significant delay in recurrence diagnosis (median 3.5 months versus 1.0 month, P < 0.0001). Uniquely, ctDNA status at the time of inconclusive CT scans predicted recurrence with positive and negative predictive values of 100%, and 75% (P = 0.0003). Serial TriMeth analysis allowed ctDNA growth rate assessment and revealed that fast ctDNA growth was associated with poor overall survival (HR: 1.6, P = 0.0052). CONCLUSIONS: Serial postoperative ctDNA analysis has a strong prognostic value and is more sensitive for recurrence detection than standard-of-care CRLM surveillance tools. Altogether, TriMeth provides several opportunities for improving postoperative surveillance of CRLM patients.

Circulating Tumor DNA in Stage III Colorectal Cancer, beyond Minimal Residual Disease Detection, toward Assessment of Adjuvant Therapy Efficacy and Clinical Behavior of Recurrences.

PURPOSE: Sensitive methods for risk stratification, monitoring therapeutic efficacy, and early relapse detection may have a major impact on treatment decisions and patient management for stage III colorectal cancer patients. Beyond assessing the predictive power of postoperative ctDNA detection, we explored the added benefits of serial analysis: assessing adjuvant chemotherapy (ACT) efficacy, early relapse detection, and ctDNA growth rates. EXPERIMENTAL DESIGN: We recruited 168 patients with stage III colorectal cancer treated with curative intent at Danish and Spanish hospitals between 2014 and 2019. To quantify ctDNA in plasma samples (n = 1,204), 16 patient-specific somatic single-nucleotide variants were profiled using multiplex-PCR, next-generation sequencing. RESULTS: Detection of ctDNA was a strong recurrence predictor postoperatively [HR = 7.0; 95% confidence interval (CI), 3.7-13.5; P < 0.001] and directly after ACT (HR = 50.76; 95% CI, 15.4-167; P < 0.001). The recurrence rate of postoperative ctDNA-positive patients treated with ACT was 80% (16/20). Only patients who cleared ctDNA permanently during ACT did not relapse. Serial ctDNA assessment after the end of treatment was similarly predictive of recurrence (HR = 50.80; 95% CI, 14.9-172; P < 0.001), and revealed two distinct rates of exponential ctDNA growth, slow (25% ctDNA-increase/month) and fast (143% ctDNA-increase/month; P < 0.001). The ctDNA growth rate was prognostic of survival (HR = 2.7; 95% CI, 1.1-6.7; P = 0.039). Serial ctDNA analysis every 3 months detected recurrence with a median lead-time of 9.8 months compared with standard-of-care computed tomography. CONCLUSIONS: Serial postoperative ctDNA analysis has a strong prognostic value and enables tumor growth rate assessment. The novel combination of ctDNA detection and growth rate assessment provides unique opportunities for guiding decision-making.See related commentary by Morris and George, p. 438.