Release Note
Recently published apps and resources
This release introduced a new set of applications, projects, and tool upgrades to the Public Apps Gallery on the Seven Bridges Platform. These additions expand support for clinical machine learning, multi-omics integration and biomedical imaging workflows, providing researchers with a broader portfolio of production-ready CWL applications for translational and bioinformatics research.
COMET Framework Applications
Three new applications have been added to support machine learning workflows built around electronic health record (EHR) data and the COMET framework. Together, these tools provide an end-to-end workflow for representation learning, model training, evaluation, and clinical prediction.
COMET Train Word2Vec
Generates patient embeddings and recurrent neural network (RNN)-ready datasets from preprocessed EHR data. This application helps researchers transform longitudinal clinical records into structured representations suitable for downstream machine learning analyses.
COMET Run Experiments
Supports training and evaluation of GRU-based EHR models and enables zero-shot testing, fine-tuning, and integrated COMET experiments. The application simplifies model development and benchmarking across clinical machine learning studies.
COMET Prediction
Performs inference using trained COMET models on new patient datasets. Researchers can apply previously trained models to generate predictions and support downstream clinical and translational research applications.
Expanded mixOmics Toolkit Portfolio
A comprehensive set of mixOmics 6.32.0 applications for multivariate and integrative omics analyses. These tools strengthen support for biomarker discovery, dimensionality reduction, data integration, and predictive modeling across multiple omics datasets.
MixOmicsPCA
Supports exploratory data analysis and dimensionality reduction, helping users identify major sources of variation within complex biological datasets.
MixOmicsPLS and sPLS
Enables modeling relationships between datasets while incorporating variable selection capabilities, making it easier to identify the most informative molecular features.
MixOmicsPLS-DA and sPLS-DA
Provides supervised classification workflows for biomarker discovery and sample discrimination across biological conditions.
MixOmicsrCCA
Facilitates the exploration of correlations between datasets, supporting the identification of biologically relevant relationships across data modalities.
MixOmicsMINT
Enables integration of multiple independent studies measuring common variables, helping researchers identify robust and reproducible molecular signatures.
DIABLO
Supports outcome-driven multi-omics integration and biomarker discovery by identifying correlated features across multiple data types associated with a phenotype or clinical outcome.
These additions significantly enhance the platform’s capabilities for integrative omics analysis, biomarker identification, and advanced statistical learning workflows.
New Public Projects
Several new public projects have been published to provide reference implementations and reproducible examples across multi-omics and imaging domains.
MixOmicsDIABLO Public Project
Demonstrates outcome-driven multi-omics integration using DIABLO to identify correlated transcriptomic and proteomic signatures associated with COVID-19 severity, based on the Harriott et al. 2025 study.
MixOmicsMINT Public Project
Showcases multivariate integrative analysis using MAQC datasets to identify reproducible molecular signatures across independent microarray studies and platforms.
nnU-Net Public Project
Provides an accessible implementation of the self-configuring nnU-Net v2 framework for biomedical image segmentation. The project supports automated preprocessing, model training, and inference for both 2D and 3D imaging datasets.
MOFA2 Public Project
Brings together MOFA2, MEFISTO, and MOFACell workflows for unsupervised multi-omics factor analysis, temporal and spatial omics integration, and single-cell or multicellular program discovery.
Resources
If you have questions or need help with any of these new resources, please contact your Velsera Seven Bridges representative or email support@velsera.com.
