Development of a blood-based cancer screening assay with a novel multivariate cancer risk score (MCRS) model by integrating shallow WGS data and plasma protein markers.

Abstract

45 Background: Despite the increase of our knowledge about carcinogenesis and recent successful development of innovative cancer drugs, cancer mortality has only slightly decreased in the past decades. Cancer early detection is probably the most cost-effective means to reduce cancer mortality as prognosis is much better when cancer is detected and treated at the early stage. Recently studies have demonstrated that blood-based mutation detection approaches may be effective to identify asymptomatic cancer patients from general population. Methods: Here we reported a novel multivariate cancer risk score (MCRS) model that interrogates shallow whole genome sequencing (WGS) data from cell-free DNA (cfDNA) and protein markers results in a single blood draw. Results: In a prospective clinical study consisting of 76 stage I-IV cancer patients and 152 normal individuals, the MCRS model demonstrated 60% sensitivity at 98.5% specificity. In order to validate clinical utility of the MCRS model in detection of cancer patients, we collated the data from a previous study of occult maternal malignancies from 1.93 million pregnant women undergoing NIPT test between 2016 and 2017. 466 out of the 639 pregnant women who tested positive for multiple chromosomal aneuploidies (MCAs) in the initial NIPTs (i.e. 0.06 - 0.1x WGS), have also underwent the protein markers test. Out of the 466 subjects, 39 maternal malignant cancer cases were diagnosed with a median follow-up of 399 days (IR, 303 - 487 days) by imaging and histology. The cancer patients presented a wide spectrum of cancer types, the most frequent being breast cancer (10 cases), liver cancer (8) and lymphoma (8), at stage II to IV. This subgroup of 466 subjects was selected as an independent validation cohort for our study. Through our new method, we analyzed shallow WGS and proteins data. The MCRS model allowed 28 of the 39 (71.8%) cancer cases to be identified with a positive predictive value of 73.7% and specificity of 97.7%. Conclusions: Taken together, these data demonstrate the MCRS model holds promise for detecting cancer in asymptomatic individuals, particularly in the populations at high risk of cancer.