Release v0.0.3

Version 0.3.0-beta of the IMPROVE software includes python tools and guidelines for structuring prediction models in a standardized format. This standardization enables the use of these models with the cross-study analysis (CSA) benchmark dataset that was compiled as part of IMPROVE. The process of restructuring community models to adhere to this standard is referred to as model curation. A key improvement in version 0.3.0-beta is the code refactoring of original model repositories that separates preprocessing, training, and inference into separate model scripts. This modular design enhances flexibility, making the code easier to maintain and integrate into a variety of workflows. This version has been employed to refactor the code structure of several community deep learning-based DRP models, covering a diverse range of model architectures, data representations, and learning schemes. These models have been integrated with Hyperparameter Optimization (HPO) and Cross-Study Analysis (CSA) workflows.

Materials for version 0.3.0-beta:

• Documentation: docs

• IMPROVE repo: lib

• Models (models that use IMPROVE version v0.0.3-beta are also tagged with v0.0.3-beta):

• GraphDRP

• LGBM

• DeepTTC

• DeepCDR

• DualGCN

• HiDRA

• IGTD

• PaccMann

• PathDSP

• tCNNS