Abstract
BACKGROUND
Current liquid-based cancer screening relies on massive deep NGS to detect rare cancer-derived genetic materials - a costly method fraught with high false-negative and false-positive rates. We aim to develop a non-NGS-centered, AI-directed liquid-based detection of GBM stem-like cells (GSCs).
METHODS
Utilizing a robust AI suite, NETZEN, we defined a common master regulatory gene network (MRGN) of the GBM state in GSCs. Since master regulators (MRs) in MRGN are developmentally restricted, their chromosomal loci are accessible in GSCs but not in normal cells. Downstream factors in MRGN are massively overexpressed in GSCs compared to normal cells. Thus, we measured the following in PBMCs from healthy controls spiked with known quantities of GSCs and patients with GBM: 1) accessibility of MR genes using transposase/transposons carrying unique barcodes to be inserted into the MR’s predetermined accessible locations, and 2) expression of downstream factors using nested qRT-PCR.
RESULTS
We characterized 10 MRs in GSCs with ≥1 promoter region that is hypomethylated and accessible (by ATACseq) in GBM/GSCs per GSE70175-92460 (19 samples) and GSE67633-96088 (14 samples), or hypermethylated and inaccessible in lymphocytes/PBMCs per GSE98837 (6) and GSE74912 (13). Using barcoded transposons, we specifically disrupted 4 MR’s accessible foci only in GSCs, not in PBMCs. We also identified 50 downstream factors with the top 20 having 3 to 5-orders-of-magnitude higher mean expression in GSCs compared to PBMCs (GSE79362-86884, 451 samples). Currently our method has a detection limit of 0.2-1 GSC per 106 PBMCs. Using the first iteration, we detected MRGN of the GBM state in blood samples of 14 of 14 GBM patients before resection, compared to none of 15 healthy donors.
CONCLUSIONS
Chromosomal accessibility and signal amplification in MRGN of GSCs provide powerful substrates for a non-NGS, low-cost, liquid-based GBM detection system with high sensitivity and specificity. Further testing and optimization are ongoing.