About

AIDrugApp v1.2.5 (https://aidrugapp.streamlit.app/) is a national award-winning, novel open-access, self-conceived project to develop a deep learning AI-based web application for drug discovery. The goal is to use AI/ML and cloud technology to develop a platform where multiple experts around the world can collaborate in drug discovery research, aiming to speed up and improve the process, especially in emergent situations like the COVID-19 pandemic. AIDrugApp built on Python3 Programming Language using libraries such as TensorFlow, Keras, scikit-learn, Matplotlib, seaborn, RDKit, pubchempy, Streamlit etc and hosted on Heroku and Streamlit platforms.

The current version 1.2.5 is for virtual screening of molecules against target proteins through Deep Neural Network (DNN) and Quantitative Structure-Activity Relationship (QSAR) based AI models and a data statistical visualization platform towards SARS-CoV-2. We have also implemented 'Custom ML Tools' into the 'AIDrugApp' framework. 'Custom ML Tools' includes four modules: 'Mol Identifier', 'DesCal', 'AutoDL', and 'Auto-Multi-ML' which provide users with free access to molecular identification using SMILES and compound names, similarity search, descriptor calculation, building of ML/DL QSAR models and their usage in predicting new data. Users can also store and use available data from the repository section for drug discovery research.

Bioactivity prediction against SARS-CoV-2

Bioactivity prediction is one the module of AIDrugApp v1.2.5 that helps to predict the bioactivities (Active/ Inactive) and pIC-50 values of users molecules of interest towards SARS-CoV-2 replicase polyprotein (RP), 3CLpro and human angiotensin converting enzymes (ACE). It is also useful for virtual screening of chemical features of molecules towards SARS-COVID-19 clinical trials (CT) bioactivities. It uses supervised deep learning algorithms with Artificial Neural Network (ANN).

Auto-Multi-ML

Automated Multiple Machine Learning (Auto-Multi-ML) is one of the Custom Machine Learning tools of AIDrugApp v1.2.3 that helps to apply and compare multiple machine learning models at the same time to select the best performing algorithm on users data. It also helps to predict target data based on user specific machine learning models.

Auto-DL

Automated Deep Learning (Auto-DL) is one of the Custom Machine Learning tools of AIDrugApp v1.2.5, where users can build the best deep learning models with neural networks for the prediction of target data.

DesCal

Molecular descriptor calculator module of AIDrugApp v1.2.5 helps to generate various molecular 2-D descriptors and fingerprints on users data. It also helps to calculate customised molecular descriptors as selected by the user on their data.

Mol_Identifier

Molecule identification module of AIDrugApp v1.2.5 helps to convert chemical names to SMILES, molecular SMILES to compound names and their 2-D structures. It also helps to detect molecular similarities on users data.

Future Work

We also conducted two case studies where large sets of molecules were screened by our app against SARS-CoV-2. Any feedback/suggestions regarding AIDrugApp are also welcome to my email ID. The future versions will include a collaborative networking platform with advanced features for many other diseases.

Email

divya.karade@gmail.com

Phone

+91 8830379882

Address

Pune, Maharashtra
India- 411008