103P - Epigenetic regulated genes enhanced fragmentomics-based model for early-stage lung cancer detection

Background

Liquid biopsy is a promising, noninvasive approach for early cancer detection. However, the extremely low portion of circulating tumor DNA (ctDNA) shedding into the blood in early lung cancer limits the screening application. Exploring the multi-layer epigenetic alterations accompanied with carcinogenesis could enhance the detection capacity of liquid biopsies. Therefore, we leveraged this cross-information to represent transcriptional and pathological characteristics of lung cancer and developed an accurate and sensitive model for early-stage lung cancer screening.

Methods

Participants with lung cancer (n=211), especially early-stage lung cancer (stage 0-I, n=145), benign nodules (n=33), and healthy volunteers (n=132) were recruited from two independent clinical centers (training center n=191, external validation center n=185). Peripheral blood cfDNA samples were simultaneously subjected to ChIP-seq, RRBS, and low-pass WGS (lpWGS). Cancer-specific synergistic effect among cell-free nucleosome H3K4me3, cfDNA methylation, and cfDNA nucleosomal deleted regions (NDRs) were analyzed to filter out Multi-Epigenetic Regulated GEnes (MERGE). A stacked ensemble machine learning model based on lpWGS was developed by integrating fragmentomic features centering around MERGE.

Results

A total of 655 MERGE were identified from a set of training cohort. Functional annotation revealed their association with transcription factors related to early lung cancer, including KLF15, SP1, and E2F families. The cfDNA fragment motifs displayed more distinct cancer-specific patterns in MERGE regions than in whole-genome. The MERGE-based integrated model was validated in an external cohort (81.7% at 0-I stage), achieving a sensitivity of 90.4% at specificity of 83.1% (AUC, 0.94), and demonstrated its high sensitivity of 93.1% at IA stage, 95.2% of minimally invasive adenocarcinoma (MIA) and 78.3% of adenocarcinoma in situ (AIS).

Conclusions

We developed a novel method by effectively enriching biologically meaningful epigenetic regulated regions, and established an integrated model for enhanced early detection of lung cancer during curable phases.

Clinical trial identification

Editorial acknowledgement

Legal entity responsible for the study

Shanghai Weihe Medical Laboratory Co. Ltd.

Funding

Shanghai Weihe Medical Laboratory Co. ltd Shanghai Weihe Medical Laboratory Co. Ltd.

Disclosure

All authors have declared no conflicts of interest.