Pan-Cancer Detection and Typing by Mining Patterns in Large Genome-Wide Cell-Free DNA Sequencing Datasets

Abstract

Background Cell-free DNA (cfDNA) analysis holds great promise for non-invasive cancer screening, diagnosis, and monitoring. We hypothesized that mining the patterns of cfDNA shallow whole-genome sequencing datasets from patients with cancer could improve cancer detection. Methods By applying unsupervised clustering and supervised machine learning on large cfDNA shallow whole-genome sequencing datasets from healthy individuals (n = 367) and patients with different hematological (n = 238) and solid malignancies (n = 320), we identified cfDNA signatures that enabled cancer detection and typing. Results Unsupervised clustering revealed cancer type-specific sub-grouping. Classification using a supervised machine learning model yielded accuracies of 96% and 65% in discriminating hematological and solid malignancies from healthy controls, respectively. The accuracy of disease type prediction was 85% and 70% for the hematological and solid cancers, respectively. The potential utility of managing a specific cancer was demonstrated by classifying benign from invasive and borderline adnexal masses with an area under the curve of 0.87 and 0.74, respectively. Conclusions This approach provides a generic analytical strategy for non-invasive pan-cancer detection and cancer type prediction.

Cell type signatures in cell-free DNA fragmentation profiles reveal disease biology

“…Overall, the reported performance of existing classifiers varies significantly even within the same disease entity depending on the specific cfDNA metric used, the size and diversity of the study population, and the algorithm employed 12 , 51 . Notably, our classifier for early-stage breast cancer shows promise compared to existing fragmentation-based classifiers that we benchmarked against (i.e., ichorCNA, Griffin) and other published classifiers 30 , 52 .…”

“…We next asked if we could detect aberrant cell type contributions in plasma cfDNA from cancer patients. We previously demonstrated that large-scale cfDNA coverage patterns vary in different cancers compared to the healthy state 30 , and we reasoned that those differences may be driven by the contribution of affected cell types. To test this, we ranked cell types in cfDNA data from individuals with colorectal cancer ( n = 16) and breast cancer ( n = 52) sequenced at 10-fold coverage and compared the rankings to those generated in healthy individuals sequenced at similar coverage and matched for sex (Supplementary Table 1 ).…”

Alleviating cell-free DNA sequencing biases with optimal transport

“…Let us denote by D 9 and D 10 the wet labs from which the samples originate (see Table 6), respectively [53] and [54]. We first built a panel of controls (reference set) using the 79 controls from domain D 9 and called CNAs in cancer cases from D 9 .…”

“…Our second data set (HEMA) focuses on haematological malignancies and is composed of 179 cases of Hodgkin lymphoma (HL), 37 of diffuse large B-cell lymphoma (DLBCL) and 22 of multiple myeloma, as well as 498 controls. Among those, 177 HL cases and 260 controls have been published in a previous study [57] and the entirety of the haematological cancer cases have been included in one of our studies (GipXplore [53]). The libraries of 242 out of the 499 controls have been prepared with the same kit as the haematological cancer cases, namely the TruSeq ChIP Library Preparation Kit (Illumina) [4].…”

Multicancer Early Detection Tests for Cancer Diagnosis

“…It is among the pioneering multicancer screening panels that leverage artificial intelligence to increase the precision of tumor marker tests, aiding in the detection of over 20 different types of cancer. 85 Additionally, a focused methylation cfDNA-based MCED test underwent five analytical validation studies using samples from 39 individuals with cancer, encompassing 12 distinct cancer types. These studies demonstrated that the MCED test had a remarkable specificity of 99.3% and accurately identified the source of cancer signals with exceptional reproducibility and consistency.…”