Flexynesis

Flexynesis is an end-to-end framework for multi-omics data integration and predictive modeling, combining data preprocessing, feature selection, Bayesian hyperparameter optimization, model training, evaluation, and interpretation within a unified workflow. It supports both deep learning and classical machine learning methods through a standardized interface, enabling classification, regression, survival analysis, multi-task learning, and cross-modality prediction from heterogeneous molecular datasets.

Flexynesis is designed with interpretability in mind. By integrating methods such as integrated gradients through Captum, it helps researchers identify informative molecular markers and better understand model predictions beyond black-box classification or regression.

Key benefits
Deep learning framework for multi-omics data integration
Supports multiple neural architectures for different prediction tasks
Flexible fusion of heterogeneous omics layers
Automated feature selection and hyperparameter optimization
Interpretability support for marker discovery using integrated gradients
Applicable to classification, regression, survival, and cross-modality prediction tasks
Continuously benchmarked on public datasets, especially in oncology
Applications
Prediction of clinical and preclinical endpoints from multi-omics data
Drug response prediction in patients and preclinical models such as cell lines and PDXs
Cancer subtype classification
Survival and outcome prediction
Cross-modality prediction between molecular data layers
Discovery of predictive biomarkers and molecular signatures
Benchmarking of deep learning approaches for biomedical multi-omics integration
Intended use

Flexynesis is intended for computational biologists, bioinformaticians, translational researchers, and machine learning researchers working with multi-omics datasets and clinically relevant endpoints. It is particularly suited for users who want to build interpretable deep learning models for biomedical prediction tasks, especially in oncology and preclinical research.

Contact:
Website https://github.com/BIMSBbioinfo/flexynesis/discussions