SAELGMDA: Identifying human microbe-disease associations based on sparse autoencoder and LightGBM

Area of Science:

• Computational biology
• Microbiome research
• Disease pathogenesis

Background:

• Identifying microbe-disease associations (MDAs) is crucial for understanding disease mechanisms and developing therapies.
• Traditional experimental methods for MDA detection are costly, time-consuming, and labor-intensive.

Purpose of the Study:

• To develop an efficient computational method, SAELGMDA, for predicting potential microbe-disease associations.
• To overcome the limitations of experimental MDA detection.

Main Methods:

• SAELGMDA integrates functional and Gaussian interaction profile kernel similarity for microbe and disease similarity computation.
• Feature vectors are created by combining similarity matrices and mapped to a low-dimensional space using a Sparse AutoEncoder.
• Light Gradient Boosting Machine is employed for classifying unknown microbe-disease pairs.

Main Results:

• SAELGMDA demonstrated superior performance compared to four state-of-the-art MDA prediction methods across HMDAD and Disbiome databases.
• The method achieved high accuracy, Matthews correlation coefficient, AUC, and AUPR values in cross-validation studies.
• SAELGMDA identified potential associations, including *Clostridium coccoides* with colorectal cancer and Sphingomonadaceae with inflammatory bowel disease.

Conclusions:

• The SAELGMDA method offers a promising computational approach for discovering novel microbe-disease associations.
• This tool can aid in identifying potential microbial biomarkers and therapeutic targets for various diseases.