cfDNA Fragments Research Articles

MATERNAL FIBROID IMPACT ON CELL FREE DNA (cfDNA) SCREENING

In cell free DNA (cfDNA) screening, a proportion of the circulating cfDNA is placental and serves as a proxy for fetal aneuploidy status. As the remaining cfDNA originates from maternal tissue, identification of unexpected maternal findings is possible. Like other tissues, fibroids can contribute cfDNA fragments to maternal plasma. ∼50% of fibroids have abnormal chromosome complements which may confound fetal aneuploidy screening.1-3 Here we investigate cfDNA data and results from 57 patients with known fibroids.

Combination of Circulating Cell-Free DNA and Positron Emission Tomography to Distinguish Non-Small Cell Lung Cancer from Tuberculosis.

Non-small cell lung cancer (NSCLC) holds high metabolic tumor burden and circulating cell-free DNA (cfDNA) levels, and the relationship between metabolic tumor burden and cfDNA in NSCLC and the underlying mechanism of their interaction therein remain poorly characterized. Our aim was to evaluate the clinical value of cfDNA and metabolic tumor burden by positron emission tomography-computed tomography (PET/CT) for NSCLC differential diagnosis from tuberculosis in patients with solitary pulmonary nodules. Metabolic tumor burden values in humans (subjects with NSCLC, subjects with tuberculosis, and healthy control subjects) and relevant mouse models were detected by preoperative 18F-fluorodeoxyglucose PET (18F-FDG PET/CT) and [3H]-2-deoxy-DG uptake, respectively. The cfDNA levels were detected by quantifying serum cfDNA fragments from the ALU (115 bp) gene using reverse transcription-polymerase chain reaction. RNA sequence was performed to determine the underlying target genes and knocked down or inhibited the target genes in vivo and in vitro to determine the mechanism therein. Metabolic tumor burden correlated with serum cfDNA levels in NSCLC subjects but not in tuberculosis subjects or healthy controls. Mouse models showed a similar phenomenon. In addition, the RNA sequence showed that glucose transporter 1 (GLU1), factor-related apoptosis ligand (FasL), caspase 8, and caspase 3 were significantly increased in NSCLC mouse tumors compared with those in tuberculosis mouse masses. Inhibiting the metabolic tumor burden by blocking or knocking down GLU1 markedly reduced the expression of FasL, the phosphorylation of caspase 8/caspase 3, and serum cfDNA levels/apoptosis percentage in vivo and in vitro. Furthermore, the use of a combination of cfDNA and metabolic tumor burden allowed better ability to distinguish NSCLC subjects from those with tuberculosis or healthy controls than either method used alone. Metabolic tumor burden promotes the formation of circulating cfDNA through GLU1-mediated apoptosis in NSCLC, and the combination of cfDNA and metabolic tumor burden could be valuable for distinguishing NSCLC from tuberculosis.

Abstract B006: Development of a NF1-MPNST-PDX liquid biopsy model using whole-genome sequencing and quantitative PCR of mouse-derived cell-free DNA

Abstract Background: Malignant Peripheral Nerve Sheath Tumors (MPNST) are aggressive, NF1-associated soft tissue sarcomas with a dismal 40% 5-year survival rate, high rates of disease relapse and metastasis, and limited treatment options. Our lab has generated a collection of patient-derived xenograft (PDX) models, which recapitulate the variety of genetic changes that occur in human NF1-MPNST. Longitudinal analysis of PDX tumor burden in preclinical studies is currently limited to imprecise tumor measurement or expensive and time-consuming imaging modalities such as MRI. These methods of analysis fail to allow for the early detection of tumor formation, the assessment of treatment-response dynamics, the formation of resistant subclones, or the detection of molecular residual disease. Liquid biopsies using plasma-derived cell-free DNA (cfDNA) have been successful in addressing these challenges in the context of multiple solid tumor types including MPNST by our group previously (Szymanski et al. 2021). However, liquid biopsy of cfDNA in the context murine model systems is quite limited, primarily due the challenge of collecting sufficient plasma for analysis. We hypothesize that we can detect and longitudinally track tumor burden in response to treatment using human-tumor-specific quantitative PCR (qPCR) and whole-genome sequencing of sequentially collected PDX-bearing murine plasma to develop a NF1-MPNST-PDX liquid biopsy model. Methods: To address this challenge, we sequentially collected ≈50 µL of murine plasma weekly from two independent groups of NF1-MPNST-PDX-bearing mice and corresponding PDX-free controls for up to six weeks. The first group consisted of six PDX-bearing mice who underwent whole-genome sequence of murine-derived plasma cfDNA and paired PDX tissue. The second group consisted of fifteen NF1-MPNST-PDX-bearing mice and five PDX-free controls; plasma cfDNA samples collected from this group underwent human-specific LINE-1 qPCR, which should only detect human DNA derived from implanted PDX and allow for longitudinal tracking of tumor burden. Results: We serially collected ≈50 µL of plasma weekly from NF1-MPNST-PDX mice with no obvious impact on mouse survival or weight. Mean total cfDNA yield from the terminally collected samples was 1.12 ± 1.02 ng per uL plasma (n=21) for PDX-bearing animals with cfDNA fragments canonically appearing between 70 and 450 bps as determined by electropherogram. Whole-genome sequencing of murine plasma cfDNA revealed broad genomic aneuploidy detectable in NF1-MPNST-PDX plasma which corresponded to alterations present in PDX tissue, including MPNST-specific chromosome 8q gain, corresponding with our earlier findings (Dehner et al. 2021). Human-specific LINE-1 qPCR showed specific enrichment of tumor-derived LINE-1 in NF1-MPNST-PDX-bearing mice compared to PDX-free mice which showed no human LINE-1 enrichment. Conclusion: Our data suggest that PDX plasma cfDNA may be an informative biomarker that can be non-invasively collected and used to track tumor burden in NF1-MPNST-PDX models. Citation Format: Paul A. Jones, Wenjia Feng, Xiaochun Zhang, Peter K. Harris, Taylor Sundby, Jeffrey J. Szymanski, Divya Srihari, Faridi Qaium, Jack F. Shern, Aadel A. Chaudhuri, Angela C. Hirbe. Development of a NF1-MPNST-PDX liquid biopsy model using whole-genome sequencing and quantitative PCR of mouse-derived cell-free DNA [abstract]. In: Proceedings of the AACR Special Conference: Sarcomas; 2022 May 9-12; Montreal, QC, Canada. Philadelphia (PA): AACR; Clin Cancer Res 2022;28(18_Suppl):Abstract nr B006.

New Perspectives on the Importance of Cell-Free DNA Biology.

Body fluids are constantly replenished with a population of genetically diverse cell-free DNA (cfDNA) fragments, representing a vast reservoir of information reflecting real-time changes in the host and metagenome. As many body fluids can be collected non-invasively in a one-off and serial fashion, this reservoir can be tapped to develop assays for the diagnosis, prognosis, and monitoring of wide-ranging pathologies, such as solid tumors, fetal genetic abnormalities, rejected organ transplants, infections, and potentially many others. The translation of cfDNA research into useful clinical tests is gaining momentum, with recent progress being driven by rapidly evolving preanalytical and analytical procedures, integrated bioinformatics, and machine learning algorithms. Yet, despite these spectacular advances, cfDNA remains a very challenging analyte due to its immense heterogeneity and fluctuation in vivo. It is increasingly recognized that high-fidelity reconstruction of the information stored in cfDNA, and in turn the development of tests that are fit for clinical roll-out, requires a much deeper understanding of both the physico-chemical features of cfDNA and the biological, physiological, lifestyle, and environmental factors that modulate it. This is a daunting task, but with significant upsides. In this review we showed how expanded knowledge on cfDNA biology and faithful reverse-engineering of cfDNA samples promises to (i) augment the sensitivity and specificity of existing cfDNA assays; (ii) expand the repertoire of disease-specific cfDNA markers, thereby leading to the development of increasingly powerful assays; (iii) reshape personal molecular medicine; and (iv) have an unprecedented impact on genetics research.

Somatic copy number alteration and fragmentation analysis in circulating tumor DNA for cancer screening and treatment monitoring in colorectal cancer patients

BackgroundAnalysis of circulating free DNA (cfDNA) is a promising tool for personalized management of colorectal cancer (CRC) patients. Untargeted cfDNA analysis using whole-genome sequencing (WGS) does not need a priori knowledge of the patient´s mutation profile.MethodsHere we established LIquid biopsy Fragmentation, Epigenetic signature and Copy Number Alteration analysis (LIFE-CNA) using WGS with ~ 6× coverage for detection of circulating tumor DNA (ctDNA) in CRC patients as a marker for CRC detection and monitoring.ResultsWe describe the analytical validity and a clinical proof-of-concept of LIFE-CNA using a total of 259 plasma samples collected from 50 patients with stage I-IV CRC and 61 healthy controls. To reliably distinguish CRC patients from healthy controls, we determined cutoffs for the detection of ctDNA based on global and regional cfDNA fragmentation patterns, transcriptionally active chromatin sites, and somatic copy number alterations. We further combined global and regional fragmentation pattern into a machine learning (ML) classifier to accurately predict ctDNA for cancer detection. By following individual patients throughout their course of disease, we show that LIFE-CNA enables the reliable prediction of response or resistance to treatment up to 3.5 months before commonly used CEA.ConclusionIn summary, we developed and validated a sensitive and cost-effective method for untargeted ctDNA detection at diagnosis as well as for treatment monitoring of all CRC patients based on genetic as well as non-genetic tumor-specific cfDNA features. Thus, once sensitivity and specificity have been externally validated, LIFE-CNA has the potential to be implemented into clinical practice. To the best of our knowledge, this is the first study to consider multiple genetic and non-genetic cfDNA features in combination with ML classifiers and to evaluate their potential in both cancer detection and treatment monitoring.Trial registration DRKS00012890.

1669P Monitoring immune checkpoint inhibition in advanced solid tumors using genome-wide cfDNA fragmentomes

The rapid identification of disease progression in patients receiving immune checkpoint inhibitors (ICIs) is challenging using methods that rely on imaging assessments. Here, we investigated genome-wide cell-free DNA (cfDNA) fragmentation profiles to molecularly detect disease progression in the phase I study of entinostat and nivolumab +/- ipilimumab in advanced solid tumors (ETCTN-9844, NCT02453620). Patients with metastatic or unresectable solid tumors received an entinostat run-in 2 weeks prior to the addition of anti-PD-1 and anti-CTLA-4 ICIs. Blood was collected at baseline (50/50 patients) and at 10 weeks after treatment initiation (27/50 patients). Low-coverage (1-2x) whole genome sequencing was performed on cfDNA from plasma samples. Genome-wide cfDNA fragmentation features were included in a Bayesian model to approximate a tumor fraction as compared to best response by RECIST v1.1. Molecular response was defined as reduction (≥ 30%) of the fragmentation based tumor fraction between the baseline and 10-week time point. 50 patients with breast (n = 28), salivary gland (n = 7), gastrointestinal (n = 6), gynecological (n = 4), sarcoma (n = 3), and lung cancer (n = 2) received a median of 4 prior lines of therapy (0-14). Response assessment using molecular analyses and best overall response by RECIST were reported at a mean of 9.3 (5-10.7) and 14.3 weeks (5-73). Molecular responders and non-responders at week 10 experienced a median PFS that was not reached (NR) and 4.5 months (HR = 5.4; 95% CI 1.7-17; p = 0.004) while patients with RECIST response and non-response at week 10 had similar PFS (median 6.4 vs. 6.7 months; HR = 1.7; 95% CI 0.54-5.3; p = 0.36). Among the subgroup of radiographic non-responders with stable disease, PFS for molecular responders was NR while for molecular non-responders it was 5.2 months (p = 0.021). There is a clinical need for non-invasive tests to identify primary or acquired resistance to ICIs, sparing patients from ineffective treatments associated with immune-related adverse events. Here, we demonstrate a cfDNA fragmentation based approach for molecular detection of disease progression that could guide future trials with ICIs.

The clinical value of circulating free tumor DNA in testicular germ cell tumor patients.

Testicular germ cell tumors (TGCT) are unique malignancies of young adult men; their biology is, however, underexplored and there has not been much progress in their treatment for decades. Circulating free tumor DNA (cfDNA) analysis represents a promising way of discovering novel diagnostic and treatment options. The study evaluates the clinical value of cfDNA detection in TGCT patients. Total cfDNA concentration and ratio of its 2 main fragments (180 and 360 bp) were evaluated by spectrophotometry, capillary electrophoresis and qPCR in peripheral blood plasma of 96 TGCT patients (173 samples) and 31 normal controls. Non-parametric tests were used for statistical analyses. The total cfDNA concentration was significantly higher in TGCT than in controls (P < 0.0001), with the highest levels at disease progression, but with no clear threshold between malignant and normal samples. Patients with positive tumor markers had higher cfDNA concentrations than those with negative markers (P = 0.01). Longer 360 bp cfDNA fragments were found in 58% of TGCT patients including almost all samples from relapse or disease progression but no normal controls (P < 0.0001). Total cfDNA levels are significantly increased in TGCT patients but without a clear threshold separating normal and tumor samples, thus total cfDNA amount itself is not a sensitive enough marker to identify or monitor TGCT. Longer cfDNA fragments have been found exclusively in a proportion of tumors and predominantly at disease progression, representing a novel potential marker for TGCT monitoring that would deserve further exploration.

916P Deep learning-based multimodal ensemble algorithm for multi-cancer detection and classification using cf-WGS

Various cell-free DNA (cfDNA) features have been investigated for their potential use in early cancer detection and cancer type classification. In this study, we developed a multimodal ensemble algorithm which integrates two independent deep learning models in which one uses regional mutational density (RMD) and mutational signature as a training feature and the other uses cfDNA fragment end motif frequency and size (FEMs) profile. Cell-free whole genome sequencing data was generated from 1,088 patients (stage I: 31.9%, II: 23.8%, III: 17.5%, IV: 11.6%, unknown: 15.2%) with breast (n=418), ovarian (n=130), pancreatic (n=97), lung (n=92), esophageal (n=142), liver (n=157), and colon cancer (n=52) and 728 healthy controls. Sequence data was produced on average of 80 million reads using Novaseq 6000. Training (60%), validation (20%) and test (20%) dataset were split stratifying cancer type and stages. RMD and FEMS models were trained using multilayer perceptron and convolutional neural networks respectively. The average value of predicted probabilities from each model was used as the final ensemble prediction. Cancer detection performance reached an accuracy of 92.8% [95% confidence interval (CI): 90.1% to 95.2%] and an AUC of 0.979 (CI: 0.964 to 0.990) in the test dataset which consist of 229 cancer patients and 146 healthy controls. Cancer detection sensitivity was 91.7% (CI: 86.0% to 96.5%) at a specificity of 95%. In multi-cancer classification, the top 1 and 2 accuracy were 78.2% (CI: 77.9% to 78.2%) and 89.5% (CI: 89.3% to 89.6%) respectively. We developed a deep learning-based multimodal ensemble algorithm for multi-cancer detection and classification. Integrating a variety of cfDNA features all together enabled highly sensitive and accurate prediction. This result shows a great potential of cfDNA WGS-based liquid biopsy in general population multi-cancer screening.

EP01.01-011 Utility of Arm-level cfDNA Fragment Size Distribution in the Early Detection of Lung Cancer and Pan Cancer

Lung cancer has the highest incidence and mortality worldwide. Early detection of lung cancer is critical for improving the prognosis of patients. cfDNA fragmentomics has shown a potential in the detection of lung cancer; however, existing predictive models lack of extensive cross-study validation, and the robustness and generalizability of the features and models should be tested and improved.

Genome-wide mutational signatures in low-coverage whole genome sequencing of cell-free DNA

Mutational signatures accumulate in somatic cells as an admixture of endogenous and exogenous processes that occur during an individual’s lifetime. Since dividing cells release cell-free DNA (cfDNA) fragments into the circulation, we hypothesize that plasma cfDNA might reflect mutational signatures. Point mutations in plasma whole genome sequencing (WGS) are challenging to identify through conventional mutation calling due to low sequencing coverage and low mutant allele fractions. In this proof of concept study of plasma WGS at 0.3–1.5x coverage from 215 patients and 227 healthy individuals, we show that both pathological and physiological mutational signatures may be identified in plasma. By applying machine learning to mutation profiles, patients with stage I-IV cancer can be distinguished from healthy individuals with an Area Under the Curve of 0.96. Interrogating mutational processes in plasma may enable earlier cancer detection, and might enable the assessment of cancer risk and etiology.

Cell-Free DNA Fragmentation Patterns in a Cancer Cell Line.

Unique bits of genetic, biological and pathological information occur in differently sized cell-free DNA (cfDNA) populations. This is a significant discovery, but much of the phenomenon remains to be explored. We investigated cfDNA fragmentation patterns in cultured human bone cancer (143B) cells using increasingly sensitive electrophoresis assays, including four automated microfluidic capillary electrophoresis assays from Agilent, i.e., DNA 1000, High Sensitivity DNA, dsDNA 915 and dsDNA 930, and an optimized manual agarose gel electrophoresis protocol. This comparison showed that (i) as the sensitivity and resolution of the sizing methods increase incrementally, additional nucleosomal multiples are revealed (hepta-nucleosomes were detectable with manual agarose gel electrophoresis), while the estimated size range of high molecular weight (HMW) cfDNA fragments narrow correspondingly; (ii) the cfDNA laddering pattern extends well beyond the 1–3 nucleosomal multiples detected by commonly used methods; and (iii) the modal size of HMW cfDNA populations is exaggerated due to the limited resolving power of electrophoresis, and instead consists of several poly-nucleosomal subpopulations that continue the series of DNA laddering. Furthermore, the most sensitive automated assay used in this study (Agilent dsDNA 930) revealed an exponential decay in the relative contribution of increasingly longer cfDNA populations. This power-law distribution suggests the involvement of a stochastic inter-nucleosomal DNA cleavage process, wherein shorter populations accumulate rapidly as they are fed by the degradation of all larger populations. This may explain why similar size profiles have historically been reported for cfDNA populations originating from different processes, such as apoptosis, necrosis, accidental cell lysis and purported active release. These results not only demonstrate the diversity of size profiles generated by different methods, but also highlight the importance of caution when drawing conclusions on the mechanisms that generate different cfDNA size populations, especially when only a single method is used for sizing.

A fetal fraction enrichment method reduces false negatives and increases test success rate of fetal chromosome aneuploidy detection in early pregnancy loss

ObjectiveWe and others have previously demonstrated that the size-selection enrichment method could remarkably improve fetal fraction (FF) in the early gestational age (GA, 12–13 weeks), suggesting that 9 or 10 weeks should not be used as a threshold for GA in size-selection noninvasive prenatal screening (NIPS). Here, we assessed whether this method was reliable for detecting fetal chromosomal aneuploidy at the earliest GA (6–8 weeks).MethodsSize-selection NIPS for fetal chromosomal aneuploidy was applied to 208 pregnancy plasma samples (102 male and 106 female fetuses), while the 169 pregnancy samples with male fetuses also underwent standard NIPS. Multivariable linear regression models were used to evaluate the association between fold-change of FF and experimental factors.ResultsThe sensitivity of the cell-free DNA (cfDNA) test in detecting aneuploidy was 100% when screened with FF enrichment, whereas the sensitivity of the same patients was only 62.5% (5/8) without FF enrichment. In the 102 pregnancy samples with male fetuses, FF increased from 6.1% to 15.7%, and the median increase in FF was 2.8-fold with enrichment. Moreover, there was a trend toward an increasing success rate of the cfDNA test from 6 to 13 weeks of gestation, especially when the test success rate reached 100% after 7 weeks with FF enrichment. Multivariate linear regression analysis demonstrated that a lower initial FF, shorter cfDNA size, increased body mass index (BMI), and later GA were all independent predictors of a higher fold-change of FF. Compared with ≤ 120 bp cfDNA fragments, the mean fold-change of FF differences was 0.820 for 121–125 bp, 0.229 for 126–130 bp, − 0.154 for 131–135 bp, − 0.525 for 136–140 bp and − 0.934 for > 140 bp (Ptrend < 0.0001), suggesting that fold-change of FF significantly decreased with cfDNA fragments > 125 bp. These results were statistically significant after adjusting for confounding factors in the models for fold-change of FF.ConclusionsThe FF enrichment method is a reasonable strategy to detect fetal chromosomal aneuploidy in early pregnancy loss with reduced false negatives and increased test success rate after 7 weeks of GA and should be recommended for patients with early pregnancy loss.

Unsupervised detection of fragment length signatures of circulating tumor DNA using non-negative matrix factorization.

Sequencing of cell-free DNA (cfDNA) is currently being used to detect cancer by searching both for mutational and non-mutational alterations. Recent work has shown that the length distribution of cfDNA fragments from a cancer patient can inform tumor load and type. Here, we propose non-negative matrix factorization (NMF) of fragment length distributions as a novel and completely unsupervised method for studying fragment length patterns in cfDNA. Using shallow whole-genome sequencing (sWGS) of cfDNA from a cohort of patients with metastatic castration-resistant prostate cancer (mCRPC), we demonstrate how NMF accurately infers the true tumor fragment length distribution as an NMF component - and that the sample weights of this component correlate with ctDNA levels (r=0.75). We further demonstrate how using several NMF components enables accurate cancer detection on data from various early stage cancers (AUC = 0.96). Finally, we show that NMF, when applied across genomic regions, can be used to discover fragment length signatures associated with open chromatin.

Improving Fetal Fraction of Noninvasive Prenatal Screening Samples Collected in EDTA-Gel Tubes Using Gel Size Selection: A Head-To-Head Comparison of Methods

The aim of this study was to compare the use of EDTA-gel blood collection tubes with and without size selection to cell-stabilizing collection tubes for remote blood sampling for noninvasive prenatal screening (NIPS). Sixty-one pregnant women at 10 to 14 weeks' gestation undergoing NIPS were recruited. Participants were phlebotomized with Streck and EDTA-gel tubes. EDTA-gel tubes were centrifuged before shipping. Libraries prepared from cell-free DNA (cfDNA) extracted from both types of tubes were sequenced on Illumina NextSeq 500, and fetal fraction was estimated using SeqFF. EDTA-gel tube libraries were size selected on agarose gel to eliminate cfDNA fragments >160 bp and resequenced. The main outcome measure was fetal fraction expressed as percentage of total cfDNA sequenced, calculated from sequence read counts (SeqFF). Streck tube samples showed an average 1% higher fetal fraction than centrifuged EDTA-gel tubes without size selection. This difference increased with temperature. When EDTA-gel samples' libraries were size selected, the mean fetal fraction increased from 7% to 13%, with no sample having fetal fraction <4%. Using EDTA-gel tubes reduces NIPS sampling cost and tube processing time in the laboratory. Also, using EDTA-gel tubes does not lead to cfDNA degradation. Size selection increases fetal fraction, reduces the number of test failures, increases NIPS clinical performance, and may be helpful in situations asking for a higher fetal fraction, such as twin pregnancies or screening for sub-chromosomal imbalances.

Cell-Free DNA as Biomarker for Sepsis by Integration of Microbial and Host Information.

Cell-free DNA (cfDNA) is emerging as a biomarker for sepsis. Previous studies have been focused mainly on identifying blood infections or simply quantifying cfDNA. We propose that by characterizing multifaceted unexplored components, cfDNA could be more informative for assessing this complex syndrome. We explored multiple aspects of cfDNA in septic and nonseptic intensive care unit (ICU) patients by metagenomic sequencing, with longitudinal measurement and integrative assessment of plasma cfDNA quantity, human cfDNA fragmentation patterns, infecting pathogens, and overall microbial composition. Septic patients had significantly increased cfDNA quantity and altered human cfDNA fragmentation pattern. Moreover, human cfDNA fragments appeared to comprise information about cellular oxidative stress and could indicate disease severity. Metagenomic sequencing was more sensitive than blood culture in detecting bacterial infections and allowed for simultaneous detection of viral pathogens. We found differences in microbial composition between septic and nonseptic patients and between survivors and nonsurvivors by 28-day mortality, both on the first day of ICU admission and across the study period. By integrating all the information into a machine learning model, we achieved improved performance in identifying sepsis and prediction of clinical outcome for ICU patients with areas under the curve of 0.992 (95% CI 0.969-1.000) and 0.802 (95% CI 0.605-0.999), respectively. We were able to diagnose sepsis and predict mortality as soon as the first day of ICU admission by integrating multifaceted cfDNA information obtained in a single metagenomic assay; this approach could provide important advantages for clinical management and for improving outcomes in ICU patients.

Abstract 5164: cell-free DNA (cfDNA) fragment size as a potential quantitative biomarker for metastatic colorectal cancer (mCRC)

Abstract Introduction: The fragment size of cfDNA derived from tumor cells are known to be shorter than that from normal cells for not yet precisely known mechanisms. We hypothesized that the fragment size of cfDNA can be utilized to reliably estimate tumor burden in patients with mCRC undergoing palliative chemotherapy. Methods: cfDNA sequencing data from mCRC patients receiving first-line palliative chemotherapy at Seoul National University Hospital (Seoul, Korea) were used for this analysis. The patients were enrolled in a main study evaluating the dynamic changes of circulating tumor DNA mutational profiles during chemotherapy. Blood samples were obtained prior to chemotherapy and after every four cycles of chemotherapy until disease progression. cfDNA sequencing data from normal individuals were used as control data. cfDNA sequencing for both cancer patients and normal individuals were performed using AlphaLiquid® 100 target capture panel (IMBdx, Inc., Seoul, South Korea). AlphaLiquid® 100 is a tumor agnostic panel consist of 106 genes, including 10 gene fusion and MSI. Based on the panel sequencing data, the genetic alterations and the fragment size of cfDNA were calculated. The fragmentation ratio was defined by the ratio of the read fragment proportion in size range P1 (100 - 155 bp) and P2 (160 - 180 bp). Results: cfDNA sequencing data from 280 plasma samples from 62 mCRC patients and 50 healthy controls were used for analysis. Compared to cfDNA from healthy controls, the cfDNA fragment sizes from mCRC patients were significantly shorter (169.585 bp vs. 173.964 bp; p &amp;lt; 0.001). Comparing the fragment sizes by the mutational status, the fragment sizes of alleles with detected somatic mutations were significantly shorter than those with reference alleles (mean fragment size 155.853 bp in alleles with somatic mutations vs. 160.613 bp in reference alleles; p &amp;lt; 0.001), but such difference was not observed with germline mutations (mean fragment size 160.911 bp in alleles with germline mutations vs. 159.889 bp in reference alleles; p = 0.992). Further, the clonality inferred from the variant allele frequency (VAF) of somatic mutation was negatively correlated with the size of the cfDNA fragment. The read fragment proportion in size range P1 was significantly associated with the clonality (r = 0.86, p &amp;lt;0.001). We calculated the mean size of DNA fragment and the fragmentation ratio with each longitudinal sample. The fragmentation ratio was found to decrease with chemotherapy in responders and re-increase at the time of resistance to chemotherapy. Conclusions: The fragmentation ratio calculated from cfDNA in mCRC patients can represent tumor burden in dynamic samples and potentially can be used as a biomarker. Citation Format: Jun-Kyu Kang, Hwang-Phill Kim, Yoojoo Lim, Su Yeon Kim, Tae-You Kim. cell-free DNA (cfDNA) fragment size as a potential quantitative biomarker for metastatic colorectal cancer (mCRC) [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr 5164.

Abstract 1912: A versatile computational pipeline for the preprocessing of cell-free DNA fragmentation data

Abstract Cell-free DNA (cfDNA) emerges as a promising liquid biopsy biomarker for cancer diagnosis and patient monitoring. Complementing mutation-based assays, cfDNA carries information about epigenetic modifications from decaying cells. This information is encoded in the shape of the cfDNA fragments. Specifically, fragments from cancer tend to be shorter than those originating from other adult cells, enabling a distinction between cancer patients and healthy individuals. Additional cfDNA features such as fragment end motifs and information on nucleosome positioning provide further insight into cancer biology. These cfDNA measures are typically inferred from low-pass whole genome sequencing and subsequent bioinformatics processing. A key bioinformatics step is the alignment of DNA sequencing reads to the reference genome, which critically depends on preprocessing steps such as read trimming and alignment filters. A good understanding of preprocessing settings is thus crucial to derive accurate information from cfDNA. We therefore investigated to what extent preprocessing choices affect cfDNA analysis. To this end, we have built a robust bioinformatics pipeline that evaluates a range of possible preprocessing settings. The pipeline is implemented in a bioinformatics workflow engine, enabling a scalable and reproducible workflow. We are currently evaluating the effect of preprocessing on global and regional fragmentation patterns, detection of nucleosome positioning and identification of fragment end motifs. The different preprocessing settings are also evaluated for their ability to distinguish cancer from control samples using cfDNA fragmentation information. We are benchmarking the pipeline using cfDNA datasets from multiple centres to ensure our findings are generalizable over different platforms and experimental procedures. Our investigations will allow us to build a versatile bioinformatics preprocessing pipeline for the analysis of cell-free DNA fragmentation data. Citation Format: Ivna Ivankovic, Zsolt Balázs, Todor Gitchev, Daniel Trejo Banos, Norbert Moldovan, Panagiotis Balermpas, Jonas Willmann, Nicolaus Andratschke, Florent Moulière, Michael Krauthammer. A versatile computational pipeline for the preprocessing of cell-free DNA fragmentation data [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr 1912.

Abstract LB112: Liquid biopsy monitoring reveals high molecular weight cfDNA fragments associated with radiotherapy dose-intensity

Abstract Introduction: Radiotherapy aims at delivering a high dose of radiation to a tumor while minimizing the dose absorbed by healthy tissues, minimizing side effects or toxicities. Radiotherapy monitoring currently largely relies on longitudinal imaging with spatial and limited functional information on therapy response. Particularly, discrimination of radiotherapy effect on the tumor versus healthy tissue can be challenging and imaging response may lag behind actual response on the tumor cell level. Liquid biopsy monitoring promises to complement imaging by providing systemic biological information about the effects of radiotherapy. Cell-free DNA (cfDNA) in the blood plasma contains genetic and epigenetic information about cell death processes, and cfDNA monitoring may thus lend itself to observing both tumor response and normal-tissue toxicity during radiotherapy. Methods: Seven polymetastatic patients, seven healthy volunteers, seven oligometastatic patients (OMD) and seven head-and neck cancer patients (H&amp;N) were recruited for the study. OMD patients received stereotactic body radiation therapy in 5 fractions of 5-10 Gy, heterogeneously prescribed, while H&amp;N underwent homogeneously 2 Gy radiation to a total dose of 70 Gy. Blood samples were collected from all patients at baseline and from patients undergoing radiotherapy twice during and twice after treatment, with samples collected up to six months after treatment. cfDNA was isolated and low-pass (7x) whole-genome sequencing was performed to examine fragment-length distribution and copy number alterations. We applied in silico size selection for 90 to 150 bp long fragments to improve copy number variant detection. Clinical measures and imaging findings were compared to cfDNA fragment analysis. Results: Samples showing high circulating tumor DNA content in copy number analysis (tMAD&amp;gt;0.03) also showed a higher proportion of short fragments (80-100 bp) compared to the samples of healthy individuals (ANOVA: p=0.0216, Tukey-pairwise: p=0.0287), whereas samples with low circulating tumor content did not (Tukey-pairwise: p=0.7540). We observed rapidly declining tumor fraction under treatment in a patient with prostate cancer and a high tMAD value at baseline. However, in the follow-up sample, residual tumor was detectable in the cfDNA copy number analysis, which was further confirmed by increasing PSA values. We observed a significant increase in the fraction of fragments between 200 and 250 bp in response to the treatment in oligometastatic patients receiving high-intensity radiotherapy, which returned to baseline in the follow-up measurements (ANOVA: p=0.0252, Tukey-pairwise: p=0.3138). A similar trend was observable (ANOVA: p=0.37) in head-and-neck patients, who received lower fractions of radiotherapy. Conclusions: Applying low-pass whole-genome cfDNA-sequencing, we were able to monitor radiotherapy response in oligometastatic and head-and-neck cancer patients. We have also observed short-term increase in the number fragments between 200 and 250 bp during treatment, which we hypothesize to be a necrotic signal. The minimal invasive nature and the relatively low cost of low-pass cfDNA-sequencing makes it a viable method for frequent monitoring, which may allow prediction of tumor response and patient outcome. Citation Format: Zsolt Balazs, Ivna Ivankovic, Panagiotis Balermpas, Todor Gitchev, Jonas Willmann, Michael Krauthammer, Nicolaus Andratschke. Liquid biopsy monitoring reveals high molecular weight cfDNA fragments associated with radiotherapy dose-intensity [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr LB112.

Abstract 5159: Towards population-scale screening of human cancer using genome-wide fragmentation profiles of cell-free DNA

Abstract Introduction: Earlier detection is a critical clinical intervention to reduce cancer-related mortality. The DELFI liquid biopsy approach (DNA evaluation of fragments for early interception) utilizing low coverage (1-2x) whole genome sequencing (WGS) to analyze cell-free DNA (cfDNA) fragmentation provides a promising avenue for cancer detection. As sequencing costs remain a barrier to adoption of liquid biopsy approaches for early detection, we evaluated WGS for DELFI using 2-channel Illumina NovaSeq sequencing as a more affordable (~7-fold cost savings) alternative to 4-channel HiSeq instruments. Methods: We performed WGS on the prospectively collected LUCAS cohort of 365 individuals at risk for lung cancer using both HiSeq and NovaSeq platforms (Mathios et. al., Nature Communications 2021). Genome-wide fragmentation was summarized in non-overlapping 5 Mb bins by ratio of short (100-150 bp) to long (151-220 bp) fragments.To measure within-sequencer repeatability, we compared fragmentation profiles of non-cancer individuals to the median non-cancer fragmentation profile by Spearman correlation. Principal component analyses were performed to assess the extent to which the sequencer explains variation of fragmentation profiles across samples.For cancer prediction, we used a penalized logistic regression model with fragmentation profiles and other genome-wide characteristics as features. Machine learning performance was assessed by cross-validation and area under the receiver operator characteristic curve (AUC). To evaluate whether we could have developed the classifier from a combination of NovaSeq with HiSeq sequenced samples, we evaluated performance trained on 90:10%, 75:25%, 50:50%, 25:75%, and 10:90% HiSeq:NovaSeq mixtures, respectively. Results: cfDNA fragmentation profiles were highly concordant among non-cancer individuals for both platforms with median correlations of 0.96 (IQR: 0.95 - 0.97) and 0.95 (IQR: 0.94 - 0.96). Visualization of fragmentation principal components did not reveal separation by sequencing platform. The DELFI approach applied to samples sequenced by NovaSeq recapitulated previously published performance measures based on HiSeq (AUC 0.90, 95% CI 0.86 - 0.94). In simulations of mixed-platform datasets, we found the same qualitative performance (AUC range: 0.893-0.902). Conclusions: cfDNA fragmentation profiles were similar between HiSeq and NovaSeq platforms, and classification accuracies from machine learning models trained on these platforms were equivalent. Our results indicate HiSeq and NovaSeq sequenced samples can be combined in models with no discernible loss in classification accuracy provided a balance of non-cancers and cancers are sequenced on both platforms. The lower cost of NovaSeq sequencing may enable wider adoption of genome-wide fragmentation-based approaches for cancer detection. Citation Format: Akshaya V. Annapragada, Dimitrios Mathios, Stephen Cristiano, Jamie E. Medina, Vilmos Adleff, Noushin Niknafs, Jacob Carey, Nic Dracopoli, Peter Bach, Jillian Phallen, Victor E. Velculescu, Robert B. Scharpf. Towards population-scale screening of human cancer using genome-wide fragmentation profiles of cell-free DNA [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr 5159.

Abstract 6371: Deep learning algorithm for multi-cancer detection and classification using cf-WGS

Abstract Purpose: Several cell-free DNA (cf-DNA) features, such as genome-wide coverage, fragment size, and fragment end motif frequency, have shown their potentials for cancer detection. In this study, we developed two independent models, GC (gross chromatin), and FEMS (fragment end motif frequency and size). Each model uses images generated from genome-wide normalized sequencing coverage and cf-DNA fragment end motif frequencies according to the different cf-DNA size profiles. Then we integrated them into a single ensemble model to improve cancer detection and multi-cancer type classification accuracy. Methods: Low depth cf-WGS data was generated from 1,396 patients (stage I: 14.9%, stage II: 35.6%, stage III: 24.9%, stage IV: 24.2%, unknown: 0.4%) with breast (n=702), liver (n=213), esophageal (n=155), ovarian (n=151), pancreatic (n=85), lung (n=53), head and neck (n=16), biliary tract (n=15), and colon cancer (n=6) and 417 healthy individuals. Samples were randomly split into training, validation, and test set stratifying cancer type and stages. Cancer types with a small number of samples (&amp;lt;20) were excluded for multi-cancer type classification. Each model was trained using a convolutional neural network, then integrated into a single ensemble model by averaging the predicted probabilities calculated from each model. Results: For cancer detection, the ensemble model achieved sensitivities of 85.2% [95% confidence interval (CI): 71.8% to 94.5%], 74.9% (CI: 68.0% to 88.0%), 73.2% (CI: 66.7% to 85.9%) at a specificity of 95%, 98% and 99% and the AUC value of 0.97(CI: 0.95-0.99) in the test dataset. By the cancer stages, sensitivity was 62.8% (CI: 48.8% to 83.7%) in stage I, 66.3% (CI: 57.7% to 82.7%) in stage II, 85.9% (CI: 77.5% to 94.4%) in stage III, and 76.1% (CI: 63.4% to 87.3%) in stage IV at 99% specificity. For multi-cancer classification, the overall accuracy of 85.1% (CI: 80.4% to 89.3%) was achieved including 6 cancer types. Conclusions: Highly sensitive and accurate deep learning model for cancer detection and multi-cancer classification was generated by combining different types of cf-DNA features. This result provides the opportunity for general population multi-cancer screening using various cf-DNA features. Citation Format: Tae-Rim Lee, Jin Mo Ahn, Joo Hyuk Sohn, Sook Ryun Park, Min Hwan Kim, Gun Min Kim, Ki-Byung Song, Eunsung Jun, Dongryul Oh, Jeong-Won Lee, Joseph J Noh, Young Sik Park, Sun-Young Kong, Sang Myung Woo, Bo Hyun Kim, Eui Kyu Chie, Hyun-Cheol Kang, Youn Jin Choi, Ki-Won Song, Jeong-Sik Byeon, Junnam Lee, Dasom Kim, Chang-Seok Ki, Eunhae Cho. Deep learning algorithm for multi-cancer detection and classification using cf-WGS [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr 6371.