A predictive machine-learning model for clinical decision-making in washed microbiota transplantation on ulcerative...

Area of Science:

• Gastroenterology and Hepatology
• Computational Biology and Bioinformatics
• Translational Medicine

Background:

• Ulcerative colitis (UC) management seeks improved clinical decision-making tools.
• Machine learning (ML) shows promise in analyzing clinical data for treatment optimization.
• Washed microbiota transplantation (WMT) is an emerging therapy for UC.

Purpose of the Study:

• To develop and validate a machine learning model for predicting the one-month clinical response to WMT in UC patients.
• To provide a novel evaluation system for clinicians and patients to optimize WMT treatment strategies.
• To identify key clinical indicators associated with WMT efficacy in UC.

Main Methods:

• A cohort of 366 UC patients undergoing WMT was retrospectively analyzed.
• Machine learning model ensembles were constructed using clinical indicators to predict treatment response.
• Internal and external validation datasets were used to assess model performance.

Main Results:

• Key predictors for WMT response included indirect bilirubin, activated antithrombin III, defecation frequency, cholinesterase, age, creatine kinase, HCO3-, and thrombin time.
• The model achieved an AUC of 0.769 ± 0.019 in internal validation and 0.614 ± 0.017 in external validation.
• An accessible online tool (https://wmtpredict.streamlit.app) was developed based on the model.

Conclusions:

• This study presents the first ML model to predict one-month WMT response in UC.
• The findings highlight the potential of ML in advancing personalized WMT strategies for gastrointestinal disorders.
• The developed model offers a valuable tool for optimizing UC treatment decisions.