search

bioFAM / MOFA2

Multi-Omics Factor Analysis
https://biofam.github.io/MOFA2/
GNU Lesser General Public License v3.0
308 stars 52 forks source link

Fast convergence of the model & highly correlated factors! #63

Closed asifemon closed 3 years ago

asifemon commented 3 years ago

Hi Folks,

I am using MOFA2 for integrating proteomics targetted, untargeted, and metabolomics datasets with a sample size of approx 500 to 800 depending on different combinations. I have normalized the datasets with quantile normalization before training. However, I am having few issues while training the model.

  1. The model is converging very fast (within 3-4 iterations) even after using convergence mode to slow.
  2. Even if I am using only 5 factors for training the model producing highly correlated factors.

As a consequence, the model can not produce any meaningful outcome (in terms of biological interpretations) so far. Any suggestions are highly appreciated. In case this is not the right place for such discussion, could you please direct me to any support forum?

warning_messages.txt

rargelaguet commented 3 years ago

Hi Asif, that’s weird. Can you send me the training output? Also, if possible, it would be useful if you could send me privately the MOFA object before the run_mofa command. Via the mofa slack group (https://biofam.github.io/MOFA2/contact.html) would be easier

Best, Ricard On 15 Feb 2021, 15:39 +0100, Asif Emon notifications@github.com, wrote:

Hi Folks, I am using MOFA2 for integrating proteomics targetted, untargeted, and metabolomics datasets with a sample size of approx 500 to 800 depending on different combinations. I have normalized the datasets with quantile normalization before training. However, I am having few issues while training the model.

  1. The model is converging very fast (within 3-4 iterations) even after using convergence mode to slow.
  2. Even if I am using only 5 factors for training the model producing highly correlated factors.

As a consequence, the model can not produce any meaningful outcome (in terms of biological interpretations) so far. Any suggestions are highly appreciated. warning_messages.txt — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub, or unsubscribe.

asifemon commented 3 years ago

solved!