dtoolAI

dtoolAI captures and manages provenance information throughout deep-learning model development to improve reproducibility and enable FAIR-compliant model and data stewardship in bioinformatics.

Key Features:

• Provenance capture: Automatically records provenance information for data, preprocessing steps, and model configurations during deep-learning model development.
• Provenance embedding: Embeds comprehensive provenance tracking mechanisms directly into model distribution artifacts.
• FAIR alignment: Structures metadata to support Findability, Accessibility, Interoperability, and Reusability of models and data.
• Model management: Provides a structured approach to generating, utilizing, and distributing deep-learning models.
• Implementation: Provided as a Python library for integration into deep-learning workflows.

Scientific Applications:

• Reproducibility in deep learning: Preserves training context to enable reproduction of model results in research settings.
• Model distribution and reuse: Facilitates sharing and reuse of deep-learning models with retained provenance and metadata.
• Bioinformatics workflows: Applies provenance-aware model management within bioinformatics research that uses deep learning.

Methodology:

Automatically capture and manage provenance information throughout model development, embed provenance tracking into workflows and model artifacts, and structure metadata to align with FAIR principles.