Tongue coating microbiota-based machine learning for diagnosing digestive system tumours.

Area of Science:

• Microbiome research
• Oncology
• Biomarker discovery

Background:

• Digestive system tumors (DSTs) are often diagnosed late due to vague symptoms.
• Early detection of DSTs is critical for improving patient outcomes.
• Non-invasive biomarkers are needed for timely DST diagnosis.

Purpose of the Study:

• To investigate the diagnostic potential of tongue coating microbiota for DSTs.
• To identify specific microbial signatures associated with DSTs.
• To evaluate the efficacy of machine learning models in classifying DSTs based on tongue microbiota.

Main Methods:

• Collected tongue coating samples from 710 DST patients and 489 healthy controls (HC).
• Analyzed microbial composition using 16S rRNA sequencing.
• Applied five machine learning algorithms, including XGBoost, to assess diagnostic performance.

Main Results:

• Significantly increased microbial diversity in the tongue coating of DST patients compared to HC.
• Identified DST-enriched genera (Alloprevotella, Prevotella) and HC-dominant taxa (Neisseria, Haemophilus, Porphyromonas).
• The XGBoost model achieved an AUC of 0.926 for DST diagnosis, outperforming other models.

Conclusions:

• Tongue coating microbiota holds promise as a non-invasive biomarker for DST detection.
• Machine learning models, particularly XGBoost, can effectively utilize microbial data for DST diagnosis.
• Further research into tongue microbiota could lead to improved early detection strategies for digestive system tumors.