search

bioFAM / MOFA

Multi-Omics Factor Analysis
GNU Lesser General Public License v3.0
230 stars 57 forks source link

AssertionError: dimensionality mismatch #59

Open wjddyd66 opened 3 years ago

wjddyd66 commented 3 years ago

My Dataset is Like this...

SNP image

Gene Expression
image

Other Dataset
image

I have two question to mofapy/run/python_template.py

First

ep.set_data(data)

When the above code is executed, the following result window appears. 

...  
Warning: columns seem to be the shared axis, transposing the data...  
...

Is this warning code to group each sample together with the same samples based on the Pandas Index?

Second

ep.train_model()

When the above code is executed, the following result window appears. 

...  
AssertionError: dimensionality mismatch
...

image

Which dimension should be matched?

For reference, in the case of other datasets (complete data, the number of samples is the same in all modalities), no error occurred.

In the case of Incomplete Data, what additional work should be done?

rargelaguet commented 3 years ago

HI @wjddyd66 , I am not sure what is going on, but please update to MOFA2 (https://github.com/bioFAM/MOFA2), as stated in the README file. The second version is much improved and this repository is no longer mantained.

See the "Getting started vignette" for Python. In addition, we now have a downstream analysis package coded in Python (https://github.com/gtca/mofax). It is less comprehensive though than the R package

Regarding SNPs, they are unlikely to be succesful in MOFA. This is because the genome does not show display many sources of variation between the features, except for population structure and related variables. This is unlike the transcriptome, which works in the form of a large number of gene regulatory networks that can vary massively across cell states, tissues, etc.