Machine learning based QSAR model for therapeutically active candidate drugs with thiazolidinedione (TZD) scaffold

Venkatesan Jayaprakash

Makale detayları

Machine learning based QSAR model for therapeutically active candidate drugs with thiazolidinedione (TZD) scaffold

Yazarlar:

Venkatesan Jayaprakash

Yayıncı:

Marmara University

Ana Başlıklar:

Pharmaceutical Chemistry

Dergi:

Journal of Research in Pharmacy Vol.28 Issue 4, pp.1135-1151

Belge Tipi:

Makale

Yayın Tarihi:

6/2025

Belge Dili:

İngilizce

Document ID (Org No.):

2001347030

pISSN:

2630-6344

Anahtar kelimeler:

Özet:

 Diabetes is a multifactorial metabolic disorder occurs due to uncontrolled persistent hyperglycaemia. The α-glucosidase enzyme plays an important role in management of diabetes. The α-glucosidase enzyme gets secreted by the brush border cells of small intestine which helps in converting maltose into glucose and thereby inhibiting the enzyme will help in lowering blood glucose level. In the present study, 100 compounds were selected having activity against α-glucosidase enzyme and they were used to build a machine learning based quantitative structure activity relationship model QSAR. All the compounds were having thiazolidinedione TZD as the common nucleus. The molecules selected were divided into training and testing datasets of 80:20 ratio for various model development. The important molecular descriptors which will affect the target were chosen using recursive feature elimination RFE algorithm. The predictive models were created using machine learning regression techniques including Support Vector Regression SVR, Random Forest Regression RFR, Decision Tree Regression DTR and Gradient Boosting Regression GBR. A comparison-based analysis was done between the various machine learning algorithms. The GBR and RFR gave the best R2 value of 0.9992 and 0.9514 for the training dataset and 0.9414 and 0.8760 for the testing dataset respectively, followed by SVR and DTR. Thus, it concludes that the four-machine learning algorithm generates a highly predictive model for the unique compounds and a superior prediction capability for building a QSAR model for α- glucosidase enzyme inhibitors.

 Diabetes is a multifactorial metabolic disorder occurs due to uncontrolled persistent hyperglycaemia. The α-glucosidase enzyme plays an important role in management of diabetes. The α-glucosidase enzyme gets secreted by the brush border cells of small intestine which helps in converting maltose into glucose and thereby inhibiting the enzyme will help in lowering blood glucose level. In the present study, 100 compounds were selected having activity against α-glucosidase enzyme and they were used to build a machine learning based quantitative structure activity relationship model QSAR. All the compounds were having thiazolidinedione TZD as the common nucleus. The molecules selected were divided into training and testing datasets of 80:20 ratio for various model development. The important molecular descriptors which will affect the target were chosen using recursive feature elimination RFE algorithm. The predictive models were created using machine learning regression techniques including Support Vector Regression SVR, Random Forest Regression RFR, Decision Tree Regression DTR and Gradient Boosting Regression GBR. A comparison-based analysis was done between the various machine learning algorithms. The GBR and RFR gave the best R2 value of 0.9992 and 0.9514 for the training dataset and 0.9414 and 0.8760 for the testing dataset respectively, followed by SVR and DTR. Thus, it concludes that the four-machine learning algorithm generates a highly predictive model for the unique compounds and a superior prediction capability for building a QSAR model for α- glucosidase enzyme inhibitors.

 Diabetes is a multifactorial metabolic disorder occurs due to uncontrolled persistent hyperglycaemia. The α-glucosidase enzyme plays an important role in management of diabetes. The α-glucosidase enzyme gets secreted by the brush border cells of small intestine which helps in converting maltose into glucose and thereby inhibiting the enzyme will help in lowering blood glucose level. In the present study, 100 compounds were selected having activity against α-glucosidase enzyme and they were used to build a machine learning based quantitative structure activity relationship model QSAR. All the compounds were having thiazolidinedione TZD as the common nucleus. The molecules selected were divided into training and testing datasets of 80:20 ratio for various model development. The important molecular descriptors which will affect the target were chosen using recursive feature elimination RFE algorithm. The predictive models were created using machine learning regression techniques including Support Vector Regression SVR, Random Forest Regression RFR, Decision Tree Regression DTR and Gradient Boosting Regression GBR. A comparison-based analysis was done between the various machine learning algorithms. The GBR and RFR gave the best R2 value of 0.9992 and 0.9514 for the training dataset and 0.9414 and 0.8760 for the testing dataset respectively, followed by SVR and DTR. Thus, it concludes that the four-machine learning algorithm generates a highly predictive model for the unique compounds and a superior prediction capability for building a QSAR model for α- glucosidase enzyme inhibitors.

 Diabetes is a multifactorial metabolic disorder occurs due to uncontrolled persistent hyperglycaemia. The α-glucosidase enzyme plays an important role in management of diabetes. The α-glucosidase enzyme gets secreted by the brush border cells of small intestine which helps in converting maltose into glucose and thereby inhibiting the enzyme will help in lowering blood glucose level. In the present study, 100 compounds were selected having activity against α-glucosidase enzyme and they were used to build a machine learning based quantitative structure activity relationship model QSAR. All the compounds were having thiazolidinedione TZD as the common nucleus. The molecules selected were divided into training and testing datasets of 80:20 ratio for various model development. The important molecular descriptors which will affect the target were chosen using recursive feature elimination RFE algorithm. The predictive models were created using machine learning regression techniques including Support Vector Regression SVR, Random Forest Regression RFR, Decision Tree Regression DTR and Gradient Boosting Regression GBR. A comparison-based analysis was done between the various machine learning algorithms. The GBR and RFR gave the best R2 value of 0.9992 and 0.9514 for the training dataset and 0.9414 and 0.8760 for the testing dataset respectively, followed by SVR and DTR. Thus, it concludes that the four-machine learning algorithm generates a highly predictive model for the unique compounds and a superior prediction capability for building a QSAR model for α- glucosidase enzyme inhibitors.

Makale detayları

Machine learning based QSAR model for therapeutically active candidate drugs with thiazolidinedione (TZD) scaffold

Yazarlar:

Venkatesan Jayaprakash

Yayıncı:

Marmara University

Ana Başlıklar:

Pharmaceutical Chemistry

Dergi:

Journal of Research in Pharmacy Vol.28 Issue 4, pp.1135-1151

Belge Tipi:

Makale

Yayın Tarihi:

6/2025

Belge Dili:

İngilizce

Document ID (Org No.):

2001347030

pISSN:

2630-6344

Machine Learning

α-glucosidase

TZD

Bioactivity

QSAR

Sayı

Makale detayları

Machine learning based QSAR model for therapeutically active candidate drugs with thiazolidinedione (TZD) scaffold

Yazarlar: Venkatesan Jayaprakash

Yayıncı: Marmara University

Ana Başlıklar: Pharmaceutical Chemistry

Dergi: Journal of Research in Pharmacy Vol.28 Issue 4, pp.1135-1151

Belge Tipi: Makale

Yayın Tarihi: 6/2025

Belge Dili: İngilizce

Document ID (Org No.): 2001347030

pISSN: 2630-6344

Machine Learning

α-glucosidase

TZD

Bioactivity

QSAR

Sayı

Yazarlar:

Venkatesan Jayaprakash

Yayıncı:

Marmara University

Ana Başlıklar:

Pharmaceutical Chemistry

Dergi:

Journal of Research in Pharmacy Vol.28 Issue 4, pp.1135-1151

Belge Tipi:

Makale

Yayın Tarihi:

6/2025

Belge Dili:

İngilizce

Document ID (Org No.):

2001347030

pISSN:

2630-6344