Novel machine-learning bioinformatics reveal distinct metabolic alterations for enhanced colorectal cancer diagnosis and monitoring.

Colorectal cancer (CRC) is the second leading cause of cancer-related mortality in the United States when considering both men and women. Colonoscopy remains the gold standard for CRC diagnosis but is invasive, costly, and requires extensive bowel preparation and sedation. Recent advancements in high throughput "omics" technologies may offer less invasive methods for CRC diagnosis through biomarker discovery. This study introduces a novel bioinformatics pipeline, PLS-ANN-DA (PANDA), combining partial least squares discriminant analysis (PLS-DA) with an advanced artificial neural network (ANN) to improve CRC diagnosis and monitor disease progression. We analyzed metabolic alterations in CRC using a metabolomics data set of 626 CRC cases and 402 healthy controls (HC). Meanwhile, complementary transcriptomic data were also analyzed and integrated to further understand CRC metabolic dysregulations. By integrating metabolomics and transcriptomics analyses and establishing the biomarker discovery pipeline PANDA, significant metabolic pathway alterations were identified between CRC patients and healthy controls, with notable upregulation of multiple pathways in CRC. Meanwhile, we observed a downregulation of specific pathways, including purine metabolism and the tricarboxylic acid (TCA) cycle, associated with advanced tumor stages. The PANDA pipeline showed promising outcomes by effectively differentiating CRC from healthy states and providing insight into metabolic shifts occurring in advanced CRC stages. Genetic mutation-associated metabolic changes were also discovered. Overall, this method has the potential for noninvasive CRC diagnostics and may serve as a valuable tool for understanding metabolic changes in cancer progression.