Herb Compounds Screening as Meningitis Inhibitor Candidates using Neural Network and Random Forest Methods Riska Aprilia(1), Mohammad Hamim Zajuli Al Faroby(2,*), Muhammad Adib Kamali(1), Muhammad Dzulfikar Fauzi(3)
(1) Information Technology Study Program, Faculty of Information Technology and Business, Institut Teknologi Telkom Surabaya, Indonesia
(2)Data Science Study Program, Faculty of Information Technology and Business, Institut Teknologi Telkom Surabaya, Indonesia
*Corresponding author: alfaroby[at]ittelkom-sby.ac.id
(3) Informatics Study Program, Faculty of Information Technology and Business, Institut Teknologi Telkom Surabaya, Indonesia
Abstract
Meningitis is an inflammation of the meninges that occurs in the protective lining of the brain and spinal cord caused by bacterial, viral, or fungal infections. This disease is difficult to recognize because it has initial symptoms like the flu where the patient has a fever and headache. Current efforts to prevent the disease by strengthening antibodies. Meanwhile, drug candidates for the treatment of this disease still have not found optimal results in reducing mortality due to meningitis. This study aims to find and analyses herbal compound candidates that might be inhibitors of meningitis. Compound data was acquired from a validated open database. The data acquired are smiles of the chemical bond structure of the compounds. In the data processing process, compound feature extraction is required by applying the concept of molecular fingerprint. The results of feature extraction are used as datasets to build classification models by applying the Multilayer Perceptron (MLP) and Random Forest algorithms. The two models are compared, and a more robust model is selected to be used as a prediction model for herbal compound search. The MLP model has a better accuracy of 0.97 compared to the Random Forest model. The results of screening using the MLP learning model obtained Symphytine, cis-Linalool oxide and 3-O-Methylcalopocarpin compounds have the highest probability compared to thousands of other herbal compounds. This candidate compound can be used as a recommendation for drug discovery to treat patients who contract Meningitis.
Keywords: Multilayer Perceptron- Random Forest- Herb Compound Screening- Meningitis- Molecular Fingerprint.
