Novartis Open Source Initiative

synadam is an R package that generates synthetic ADaM clinical trial datasets from real data using a glimpse-then-simulate approach. It preserves statistical properties and data structure while removing patient identifiers, enabling safe data sharing for pharmaceutical research and regulatory collaboration.

Surrogate models for compound ADME property prediction, including model training and evaluation

A Python library for benchmarking uncertainty estimation and quantification methods for Machine Learning models predictions.

Multi-task learning with orthogonality / disjoint-support constraints (R package)

Deep survival analysis made easy

A Python framework for evaluating and comparing multi-objective optimization (MOO) algorithms in molecular design and drug discovery.

Implementation of caMPO (Correlation-aware Multiparameter Optimization), a method for deriving data-driven non-negative MPO weights from property correlation structure in drug discovery.

Public gallery of BR Open Source projects

Applied modelling in drug development: flexible Bayesian regression modelling in Stan via brms

QM-ATLAS is a python package that orchestrates QM calculations. Based on thin wrappers of external software, more complex modules are built, exposing single calculation steps as well as multi-step workflows. The repository provides a flexible command-line interface that allows to define complex workflows in single configuration files.