Background: Colorectal Cancer (CRC) is the second leading cause of cancer mortality worldwide although highly curable if detected early. An accurate liquid biopsy test for early cancer and precancerous lesion, detection is key to promote prevention and reduce mortality. We previously developed a liquid biopsy test based on immune cells response to the tumor, commercialised under the name of Colox, with best-in-class clinical performances for AA detection (52% sensitivity and 92% specificity) (Ciarloni L et al, Clin Cancer Res, 2016). An improved second generation of the test, leveraging transcriptome profiling and advanced machine learning tools, is under development and first results are presented. Methods: Prospective peripheral whole-blood (PAXgene) samples from subjects diagnosed with CRC, advanced adenoma (AA), non-advanced adenoma, other types of cancer as well as controls without colorectal neoplasia (CON) were divided into discovery and validation sets. Transcriptome profiles were generated by RNA-seq and gene expression signatures identified leveraging Novigenix’s proprietary LITOseek platform, which integrates several advanced Machine Learning (ML) methods into a ranking systems. Biological functional analysis was performed by over-representation (ORA), gene set enrichment (GSEA) and gene network analyses (STRINGdb). Results: Significant changes in the whole blood transcriptome profile of AA compared to CON were identified. Functional analysis highlighted upregulation of fatty acid derivative biosynthesis, interferon signaling, tryptophan catabolism to kynurenine, and downregulation of DNA repair in AA blood samples compared to CON. A novel gene classifier was generated, showing for AA detection 71% sensitivity, 94% specificity and an AUC of 74% by cross-validation on the discovery set. Validation of the gene classifier in an independent set is currently ongoing and will be presented at the congress. Conclusions: Capturing immune-related information using whole blood trascriptome profiling and applying cutting-edge machine learning technologies demonstrated to be valuable for identifying gene signatures for advanced adenoma detection. This differentiated solution has demonstrated best-in-class potential to significantly improve patient outcomes by detecting precancerous lesions with a simple non-invasive blood test.