Chat Now

Drug Discovery Tools & APIs

The MediAI Drug Discovery Assistant leverages several powerful tools and APIs to provide accurate and valuable drug modification suggestions.

PubChem API

Chemical Database
Implemented

PubChem is a database of chemical molecules and their activities against biological assays. The API provides access to chemical structures, properties, and more.

Learn more

Integration Notes

Currently used to fetch SMILES notations and basic compound properties. Integration is via REST API calls.

LLM Explanations

AI/NLP
Implemented

Generates detailed scientific explanations of drug modifications using large language models, explaining mechanisms and potential benefits or drawbacks.

Integration Notes

Integrated via DeepSeek and OpenAI APIs, with automatic fallback between providers. Converts structured chemical data into accessible scientific explanations.

RDKit

Cheminformatics
Implemented

An open-source cheminformatics toolkit that provides functionality for working with chemical structures, substructure searching, molecular rendering, and more.

Learn more

Integration Notes

Will require Python backend. Planned for analyzing and modifying chemical structures based on SMILES formats.

DeepChem

Machine Learning
Implemented

DeepChem provides a high quality open-source toolchain for deep learning in drug discovery, materials science, and other fields of chemistry.

Learn more

Integration Notes

Will require Python backend with GPU support for optimal performance. Planned for molecular property prediction.

IBM RXN for Chemistry

AI Chemistry
Planned

An AI system that predicts the outcomes of chemical reactions and helps with retrosynthesis planning.

Learn more

Integration Notes

Requires API key. Planned for evaluating synthetic feasibility of proposed molecular modifications.

SwissADME

Pharmacokinetics
Planned

Free web tool to evaluate pharmacokinetics, druglikeness and medicinal chemistry friendliness of small molecules.

Learn more

Integration Notes

No official API. Would require web scraping or custom adapter. Planned for evaluating drug-likeness of modified molecules.

MedCAT

NLP/Medical
Implemented

Medical Concept Annotation Tool for extracting medical concepts from text and linking them to biomedical ontologies.

Learn more

Integration Notes

Requires Python backend. Planned for advanced drug name extraction and understanding modification goals from user queries.

SMILES/InChI Format

Chemical Representation
Implemented

Simplified Molecular Input Line Entry System (SMILES) and InChI are textual representations of chemical structures.

Integration Notes

Currently implemented to represent molecular structures. Used as the foundation for all molecular manipulations.

Implementation Architecture

Current Implementation

  1. User enters a query about drug modifications
  2. Backend extracts drug names from the query (simple keyword matching)
  3. System retrieves predefined modification suggestions for known drugs
  4. If no predefined modification exists, system attempts to fetch the drug's SMILES format from PubChem
  5. The system returns a detailed response about suggested modifications
  6. An AI-generated scientific explanation of the modification is provided using LLMs

Full Implementation (Planned)

  1. User enters a query about drug modifications
  2. MedCAT processes the query to extract drug names and modification goals (reduced toxicity, improved bioavailability, etc.)
  3. System retrieves drug information and SMILES from PubChem
  4. RDKit analyzes the molecular structure and suggests potential modifications
  5. IBM RXN evaluates the synthetic feasibility of the modifications
  6. DeepChem predicts properties of the modified molecules
  7. SwissADME evaluates pharmacokinetic properties
  8. System returns comprehensive analysis with visualizations and property predictions
  9. AI generates a detailed scientific explanation of the modification with mechanistic insights

APIs Required

  • PubChem API for drug information and SMILES (free, implemented)
  • DeepSeek/OpenAI API for generating scientific explanations (requires API key, implemented)
  • IBM RXN API for reaction prediction (free for basic usage)
  • SwissADME for property prediction (free web service, may require custom adapter)
  • Python backend for RDKit and DeepChem integration (both are free, open-source libraries)