AI-Driven Biomarker Identifies High-Risk Liver Cancer Patients Before Tumor Formation

Researchers at the RIKEN Center for Integrative Medical Sciences in Japan have unveiled a novel predictive tool for hepatocellular carcinoma (HCC), the most lethal subtype of liver cancer. Published recently in the Proceedings of the National Academy of Sciences, the research elucidates the pivotal role of the MYCN protein in driving liver tumorigenesis and introduces an innovative machine-learning algorithm capable of forecasting cancer risk by decoding the tumor-promoting microenvironments in the liver before malignancy even manifests.

Liver cancer continues to represent a formidable global health challenge, claiming over 800,000 lives annually due to its asymptomatic progression and high rates of recurrence, which linger between 70 and 80%. Current diagnostic paradigms are often inadequate for early detection, emphasizing the urgent need for biomarkers that can identify patients at elevated risk for cancer development prior to tumor formation.

The team sought to fill this gap by focusing on the MYCN gene, a member of the MYC family of proto-oncogenes known to be implicated in various cancers but whose function in liver pathophysiology was not fully understood. To robustly investigate MYCN's role in liver tumorigenesis, the researchers employed a sophisticated genetic engineering approach involving hydrodynamic tail vein injection to insert the MYCN transposon directly into the genome of mouse hepatocytes. This genetic manipulation created a mouse model with enforced overexpression of MYCN within liver tissue.

Strikingly, when MYCN was co-expressed with a constitutively active form of AKT—a kinase frequently associated with cellular growth and survival—an astounding 72% of these genetically modified mice developed liver tumors within 50 days, recapitulating many histopathological and molecular features of human HCC. Control groups expressing either gene alone did not develop tumors, underscoring the synergistic oncogenic potential of MYCN alongside AKT activation.

The team leveraged spatial transcriptomics, an avant-garde technique that maps gene expression within the histological architecture of tissue sections. This method permits unparalleled resolution in understanding where and when gene activation changes occur during tumor evolution. Applying this technology to the murine metabolic dysfunction-associated liver cancer model, the researchers tracked temporal and spatial shifts in gene expression linked to regions exhibiting elevated MYCN levels even in tumor-free liver areas.

Their spatial transcriptomics analysis identified a distinctive cluster of 167 genes differentially expressed within non-tumorous liver tissue exhibiting high MYCN—termed the "MYCN niche." This microenvironment appears to prime hepatocytes and surrounding cells for malignant transformation, acting as a permissive zone for tumor initiation. The profound biological insights gleaned from this gene signature underscore the pre-tumoral changes that herald cancer onset, opening avenues for interception before disease progression.

Capitalizing on these findings, the team developed a sophisticated machine-learning model trained on the spatial transcriptomic data. This algorithm quantifies the presence of the MYCN niche by analyzing gene-expression patterns characteristic of this pre-neoplastic milieu. Remarkably, the model achieves a 93% accuracy in distinguishing MYCN niche-positive regions, effectively serving as a computational biomarker predictive of liver cancer risk.

The team applied the MYCN niche score to human HCC datasets. Patients whose non-tumor liver tissues exhibited higher MYCN niche scores were found to have increased rates of tumor recurrence and poorer overall outcomes, highlighting the potential of this biomarker in prognostication. Intriguingly, this correlation was more pronounced when the scoring was based on non-cancerous tissue, reinforcing the concept that the tumor microenvironment—prior to overt cancer—is critical in determining patient prognosis.

This study represents a paradigm shift by marrying cutting-edge spatial transcriptomics with artificial intelligence to unveil the preclinical biological states that predispose to cancer initiation. The MYCN niche score exemplifies a new class of spatial biomarkers that transcend traditional diagnostic markers by scrutinizing the microenvironmental context that fosters disease emergence. The research team aspires to delve deeper into the biological mechanisms underpinning the MYCN niche.