MOFA2 1.7.3 (latest):

• mofapy2 has been updated to version 0.6.7
• Updated basilisk to use mofapy2 0.6.7
• Minor changes

MOFA2 1.7.2:

• Fix string size problems with HDF5
• Updated basilisk to use the most updated numpy, scipy, h5py libraries
• Spike and slab on the weights is set to False by default
• use_float32 in data options is set to True by default (speeds up training by 2x)
• mofapy2 has been updated to version 0.6.5

MOFA2 1.3.4:

• Added a more flexible alignment option in MEFISTO to align distinct sets of groups instead of individual groups
• mofapy2 has been updated to version 0.6.4

MOFA2 1.2.0:

• Improve interoperability with Seurat and SingleCellExperiment
• MOFA factors can be saved to a Seurat object using add_mofa_factors_to_seurat
• Automatically extract metadata from Seurat and SingleCellExperiment objects
• Improve memory usage and training time by optionally replacing float64 arrays with float32 arrays (specified as data_options in the prepare_mofa step)
• mofapy2 has been updated to version 0.6.0 and now it has its own repository

MOFA2 1.1.7:

• Added MEFISTO into MOFA2
• MOFA2 Package available via Bioconductor
• Improving Python interface with basilisk
• Sample metadata can be incorporated to the MOFAobject before and after training using the samples_metadata function

The MOFA package was deprecated and replaced by MOFA2 1.0.0:
The following new features are now available:

• Multi-group functionality: intuitively, this functionality breaks the assumption of independent samples and allows for are predefined groups of samples (i.e. different conditions, batches, cohorts, etc.). Importantly, the model is not focused on capturing the differential changes between the groups (you will find no factors that separate the groups). The aim of the multi-group framework is to discover which sources of variability are shared between the different groups and which ones are exclusive to a single group. Note that this functionality is optional.

• Improved downstream visualisations: see the documentation in the Tutorials section for a comprehensive list of available functions.

• No need for model selection: The old package used random parameter initialisation, which led to (slightly) different solutions depending on the initial conditions. In MOFA2 we initialise the factors using Principal Component Analysis on the concatenated data set, and the weights are initialised to zero. If using standard variational inference (not stochastic) this removes the randomness in the training algorithm.

• Speed: the training procedure is now 2-3x faster in standard CPUs.

• GPU support: the training procedure can be massively accelerated using GPUs. For this you have to install and configure the CuPy package.