Identify Diabetes-related Targets based on ForgeNet_GPC

Area of Science:

• Computational biology
• Bioinformatics
• Drug discovery

Background:

• Identifying therapeutic targets is crucial for efficient drug development.
• Polygenic diseases like diabetes result from complex gene-environment interactions.

Purpose of the Study:

• Propose a novel disease target identification algorithm based on protein recognition.
• Develop an accurate computational method for classifying disease-related proteins.

Main Methods:

• Collected diabetes-related and unrelated targets from literature.
• Constructed a protein dataset using transcribed proteins.
• Employed six feature extraction algorithms (AAC, CKSAAGP, DDE, DPC, GAAP, TPC) to generate feature vectors.
• Developed a novel classifier, forgeNet_GPC, integrating forgeNet and Gaussian Process Classifier (GPC).

Main Results:

• The proposed forgeNet_GPC classifier effectively classifies proteins.
• forgeNet selects important features, while GPC handles classification.
• forgeNet_GPC demonstrated superior performance compared to 22 existing classifiers.

Conclusions:

• The forgeNet_GPC algorithm offers a robust approach for disease target identification.
• The method shows significant improvements in classification accuracy metrics (ROC-AUC, PR-AUC, MCC, Youden Index, Kappa).
• This computational tool can accelerate the discovery of new therapeutic targets for polygenic diseases.