Standardization of shotgun metagenomics parameters for pks-positive enterobacteria detection in human fecal samples: Receiver operating characteristic analysis using polymerase chain reaction as reference standard

Abstract

Enterobacteria harboring the polyketide synthase ( pks ) island ( clbA – clbS ), which encodes colibactin ( pks⁺ enterobacteria), is a putative risk factor for colorectal cancer. Polymerase chain reaction (PCR) is the current standard for detection. Primer-independent shotgun metagenomic sequencing (SMS) is increasingly used; however, thresholds and analysis parameters for identifying pks⁺ enterobacteria remain unstandardized. This study aimed to evaluate how accurately SMS can reproduce PCR-defined pks⁺ status and identify practical cutoffs by analyzing fecal samples from 223 healthy adults in a Japanese cohort. PCR identified 60/223 (26.9%) participants as pks⁺ . SMS detected at least one clb gene in 85/223 (38.1%) at ≥ 90% identity; 49 of these agreed with PCR positives. Among the single-gene rules, clbB yielded the best discrimination (area under the curve (AUC) = 0.826) across 70–90% identity. The total-read rule across 19 clb genes performed best overall (AUC = 0.838), with higher AUCs above the 60% threshold (all p ≤ 0.045). The rule requiring detection of at least eight clb genes was inferior (AUC = 0.734–0.773). Detection performance depended on gene length; longer clb genes (> 1,500 amino acids) achieved higher accuracy (AUC ≥ 0.800). Optimized SMS can reproduce PCR-based detection of pks⁺ enterobacteria in feces with high accuracy. clbB and a cluster-level rule aggregating reads across clbA – clbS are effective, although tightening the sequence-identity threshold did not consistently improve performance. The standardized criteria proposed here enable cross-study comparability and may accelerate epidemiologic assessment of pks⁺ enterobacteria and its clinical translation for colorectal cancer prevention.