kuntalsaha05/egfr-drugdiscovery-xai

How kuntalsaha05/egfr-drugdiscovery-xai works

This is a Streamlit assistant for early virtual screening of compounds against human EGFR. A user submits a SMILES structure; the application validates and profiles it, predicts activity with the trained RF/GB ensemble, and presents structural views, SHAP explanations, drug-likeness results, and similarity to EGFR therapies. It can also produce PDF dossiers. A separate data-preparation path harvests ChEMBL bioactivity records and trains the models used for inference. Edges reflect the excerpts and README; some UI-to-report wiring is documented but not visible in the sampled app excerpt.

Main entry points: app.py Streamlit app · chem_engine.py SMILES and chemoinformatics · train_model.py Model training and evaluation · ai_model.py Activity inference

Components 14

Screening experience

Streamlit app DrugDiscoveryAI/app.py

Molecular analysis

SMILES and chemoinformatics DrugDiscoveryAI/chem_engine.py
2D/3D structure views DrugDiscoveryAI/chem_engine.py
Drug-likeness scorecard DrugDiscoveryAI/chem_engine.py
EGFR drug similarity DrugDiscoveryAI/similarity_engine.py

Data and models

Activity inference DrugDiscoveryAI/ai_model.py
Trained model artifacts DrugDiscoveryAI/models
Model training and evaluation DrugDiscoveryAI/train_model.py
Curated EGFR dataset DrugDiscoveryAI/data/egfr_chembl203_bioactivity.csv
Bioactivity data processing DrugDiscoveryAI/fetch_data.py

Interpretation and outputs

SHAP attribution DrugDiscoveryAI/explainability.py
PDF screening dossier DrugDiscoveryAI/report_generator.py

Other components

Researcher
ChEMBL service

Connections 17

Researcher → Streamlit app: submits SMILES README.md
Streamlit app → SMILES and chemoinformatics: validates and profiles DrugDiscoveryAI/app.py
Streamlit app → 2D/3D structure views: requests structure views DrugDiscoveryAI/app.py
Streamlit app → Drug-likeness scorecard: shows scorecard README.md
Streamlit app → Activity inference: requests prediction DrugDiscoveryAI/app.py
Activity inference → SMILES and chemoinformatics: featurizes molecule DrugDiscoveryAI/ai_model.py
Trained model artifacts → Activity inference: loads models DrugDiscoveryAI/ai_model.py
Streamlit app → SHAP attribution: requests explanation DrugDiscoveryAI/app.py
SHAP attribution → SMILES and chemoinformatics: maps atoms and renders DrugDiscoveryAI/explainability.py
Streamlit app → EGFR drug similarity: searches analogues DrugDiscoveryAI/app.py
Streamlit app → PDF screening dossier: generates dossier README.md
PDF screening dossier → SMILES and chemoinformatics: renders molecule and scorecard DrugDiscoveryAI/report_generator.py
ChEMBL service → Bioactivity data processing: provides bioactivity records DrugDiscoveryAI/fetch_data.py
Bioactivity data processing → Curated EGFR dataset: writes curated records DrugDiscoveryAI/fetch_data.py
Curated EGFR dataset → Model training and evaluation: supplies training data DrugDiscoveryAI/train_model.py
Model training and evaluation → SMILES and chemoinformatics: extracts features DrugDiscoveryAI/train_model.py
Model training and evaluation → Trained model artifacts: saves models and metrics DrugDiscoveryAI/train_model.py

More Python diagrams: silx-kit/pyfai · studylions/studylion · superxslam/tartanimu · swaystar123/speedrundit · touero/easier-docker

Drawn by GitDiagram on September 30, 2026 from the repository's file tree, README and source files. Read as Markdown ·