Detection of Colorectal Carcinoma Based on Microbiota Analysis Using Generalized Regression Neural Networks and...

Area of Science:

• Microbiology
• Bioinformatics
• Oncology

Background:

• Colorectal cancer (CRC) screening methods require improvement for accuracy and early detection.
• Gut microbiota composition is increasingly recognized as a potential biomarker for various diseases, including cancer.
• Identifying specific microbial species associated with CRC can lead to novel diagnostic tools.

Purpose of the Study:

• To develop an optimal classifier for colorectal cancer (CRC) detection using gut microbiota.
• To identify a novel nonlinear feature selection method for pinpointing discriminative microbial species for CRC.
• To assess the potential of gut microbiota as a non-invasive screening tool for CRC.

Main Methods:

• Modeled intestinal microflora from fecal samples of 141 patients using general regression neural networks (GRNNs).
• Employed a proposed nonlinear feature selection method to identify key microbial species.
• Validated the model's performance on independent cohorts from America and Austria.

Main Results:

• The proposed model achieved high accuracy (AUC = 0.911) in CRC detection.
• Identified Clostridium scindens and Bifidobacterium angulatum as indicators of healthy gut flora, reduced in CRC.
• Found Fusobacterium gonidiaformans to be closely correlated with CRC; adenoma detection was less discriminatory.

Conclusions:

• The novel feature selection method combined with GRNN shows potential as an accurate CRC detection tool.
• Gut microbiota changes associated with CRC appear more pronounced in advanced stages than in early adenoma.
• Integrating the model with fecal occult blood test (FOBT) did not significantly alter detection accuracy (AUC = 0.915).