Canonical text (the exact bytes fingerprinted)

TRUVACE RECORD VERSION
record: TRV-2026-2006
version: 1
kind: certified
reason: Certified into the record
timestamp: 2026-10-20T10:42:37.941684Z
status: published
lens: g_space
sector: health
headline: Deploying machine learning and metagenomics for predicting gut microbial drug biotransformation in colorectal cancer
dek: Colorectal cancer (CRC) is one of the leading causes of death globally. Chemotherapy is considered as one of the standard treatment option for CRC. Although effective in many cases, drugs used in chemotherapy are often associated with adverse toxicity. Inter-individual variations in drug response are often attributed to pharmacogenomic associations, and current evidence suggests involvement of gut microbes as a contributor to drug response variations. While several studies have implemented machine learning to pr…
gain_title: A machine learning tool combining enzyme prediction with species abundance in colorectal cancer metagenomes can identify gut microbial enzymes and species able to biotransform drugs, which may help explain variable chemotherapy response.
problem_title: (none)
trace_subject: (none)
gain_reading: A machine learning tool combining enzyme prediction with species abundance in colorectal cancer metagenomes can identify gut microbial enzymes and species able to biotransform drugs, which may help explain variable chemotherapy response.
gain_evidence: a tool that predicts gut microbial enzymes and species capable of biotransforming drugs
problem_reading: (none)
problem_evidence: (none)
quick_read: Researchers built CRCGMMeta Predictor, a machine learning tool that identifies which gut microbial enzymes and bacterial species can chemically transform drugs. It links enzyme prediction to species abundance in colorectal cancer metagenomes drawn from eight geographically diverse studies, rather than relying only on substrate and genome data.

The work matters because chemotherapy for colorectal cancer varies in effect and toxicity between patients, and gut microbes are suspected contributors. The tool is an early computational step toward microbiome-informed drug response research, but the source reports only model performance and literature comparison, not outcomes in patients.
limitation: The cross-validation accuracy was lower and more variable than the headline figures, and the source describes only computational validation, so effects on patient treatment are not shown.
tag: Evidence-backed gain
key_points: The model was trained on 437,079 species-specific enzyme records with 3457 substrate and protein-encoding features. | A random forest model scored 99% on accuracy, precision, recall, F1 and ROC-AUC during evaluation. | A blind validation set reached 97% accuracy, while 10-fold cross-validation gave 0.831 ± 0.068. | The tool was assessed on fluoropyrimidine drugs against literature evidence and existing computational tools.
rundown: Colorectal cancer is described as a leading cause of death globally, and chemotherapy is a standard treatment that is often linked to adverse toxicity. Differences between individuals in drug response are usually attributed to pharmacogenomics, though evidence suggests gut microbes also contribute.

Earlier machine learning work predicted drug responses mediated by gut microbial enzymes mainly from substrate information and bacterial genome data. The new tool adds species-level abundance from colorectal cancer metagenomes collected in eight geographically diverse studies.

After comparing several machine learning models, the authors chose random forest to build the tool. They then checked its predictions for fluoropyrimidine drugs against enzymes supported by literature and those flagged by other computational tools.
sources:
- peer_reviewed | Computational Biology and Chemistry | https://doi.org/10.1016/j.compbiolchem.2026.109495 | 2026-10-13
prev: 0000000000000000000000000000000000000000000000000000000000000000

sha256

76e4e11cf3553d4c58de53a7526a79c857437cd4905ebc99a5ef437477813c4d

previous

0000000000000000000000000000000000000000000000000000000000000000