mofa-mofax

Use mofax.mofa_model("model.hdf5") to open the file in readonly mode, then call get_factors(df=True), get_weights(df=True), and get_r2() to retrieve factors, weights, and variance explained as pandas DataFrames. Call model.close() when finished to release the HDF5 handle.

mofax provides seaborn-backed wrappers such as plot_factors, plot_weights, plot_weights_heatmap, and plot_r2 that mirror the MOFA2 R plot family. Pass the loaded mofa_model object plus factor and view names to generate the figures directly.

No, mofax is strictly read-only and cannot fit models. Train with mofapy2 in Python or MOFA2 in R first, then use mofax to inspect the resulting HDF5 file.

Weights are stored per view, so get_weights(df=True) returns one DataFrame per view keyed by view name. Concatenate them with pd.concat(w_dict, names=["view", "feature"]) if a single combined frame is needed.

mofapy2 does not always serialize samples_metadata into the HDF5 file. Build the DataFrame yourself and assign it via model.metadata = my_df, then color plots by those columns.

Avoid mofax when you need to fit a model, and do not expect a biplot since Z and W have independent scalings and cannot share axes. Also remember it holds an open readonly HDF5 handle that must be closed explicitly.