Abstract Molecular classification of colorectal cancer (CRC) based on gene expression profiling of tumor samples is known to be heavily affected by transcripts of stromal origin. As a consequence, current CRC transcriptional subtypes reflect an admixture of cancer cell-intrinsic traits and tumor microenvironment features. Whether selective analysis of the cancer cell transcriptome could improve CRC subtyping, remains an open issue. In patient-derived xenografts (PDXs), human transcripts only originate from cancer cells, because stromal transcripts are of mouse origin. We therefore assessed cancer-cell intrinsic transcriptional features of CRC by generating human-specific expression profiles of 515 PDXs from 244 CRC patients, and performing unsupervised class discovery. We identified five “CRC intrinsic subtypes” (CRIS A-E) only partially overlapping with the current ones, and robustly enriched for distinct molecular, functional and phenotypic traits: (i) CRIS-A: mucinous, glycolytic, CIMP, and enriched for microsatellite instability or mutations in KRAS; (ii) CRIS-B: marked TGF-β pathway activity, epithelial-mesenchymal transition and poor prognosis; (iii) CRIS-C: MSS, elevated EGFR signaling and sensitivity to EGFR-targeted treatments; (iv) CRIS-D: MSS, WNT activation, and IGF2 overexpression and amplification; (v) CRIS-E: MSS, Paneth cell-like phenotype and higher frequency of TP53 mutation. CRIS subtypes successfully categorized independent sets of primary and metastatic CRCs and cell lines, for a total of over 3000 samples profiled by microarrays and RNAseq. The new subtypes displayed unprecedented predictive and prognostic performances, indipendent from known markers including stromal signatures, whose integration with CRIS further enhanced prognostic significance. Citation Format: Claudio Isella, Sara E. Bellomo, Francesco Brundu, Francesco Galimi, Elisa Ficarra, Luigi Marchionni, Livio Trusolino, Andrea Bertotti, Enzo Medico. Cancer cell-selective transcriptome analysis reveals new colorectal cancer molecular subtypes with improved biological resolution and superior predictive and prognostic performance [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2017; 2017 Apr 1-5; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2017;77(13 Suppl):Abstract nr 1742. doi:10.1158/1538-7445.AM2017-1742