New algorithm to predict colorectal cancer based on fecal volatile organic compounds profile.

In this study, an algorithm designed for the analysis of fecal samples for colorectal cancer diagnostics, utilizing the data from the advanced technique of thermal-desorption-gas chromatography-mass spectrometry (TD-GC-MS), is constructed. The algorithm performs a comprehensive analysis across the entire spectral range to identify compound patterns for differentiating among three distinct health states: colorectal cancer, colorectal adenomas and controls with normal colonoscopy. The algorithm underwent a rigorous optimization process, resulting in a sensitivity and specificity of 100 %, effectively eliminating both false positives and false negatives. During the validation phase, the algorithm demonstrated remarkable performance, with sensitivity ranging from 74 % to 68 %, specificity ranging from 58 % to 52 %, and accuracy 66 %-62 % (range across twenty randomized train-test splits). Notably, in the context of polyp samples, the algorithm obtained a sensitivity range from 54 % to 50 %, even when trained with data from only healthy individuals (i.e., controls) and cancer patients. Moreover, a detailed table of compounds and their probabilities of occurrence in cancer, adenomas, and healthy samples is provided, offering insight into the interpretability of the algorithm. This qualitative approach signals a significant advancement in diagnostic precision and promises to enhance early detection of colorectal cancer, marking a substantial contribution to the field.