search

kordk / torch-ecpg

(GPU accelerated) eCpG mapper
BSD 3-Clause "New" or "Revised" License
2 stars 0 forks source link

Running MLR module without covariates #40

Open harrsha4 opened 1 month ago

harrsha4 commented 1 month ago

Thank you for this wonderful tool. The regression command usually requires covariates to run the model. Is there an option to exclude the covariates when running the regression?

kordk commented 1 month ago

Thank you for your feedback. Currently, the mlr method does require a covariate matrix. We will need to add the option for the mlr method to run without a covariate matrix.

In the meantime, one option is to run the analysis using the correlation method (corr), which will also test for a relationship between gene expression and methylation levels.

tecpg run corr --help

Usage: tecpg run corr [OPTIONS]

  Calculate the pearson correlation coefficient.

  Calculate the pearson correlation coefficient with methylation and gene
  expression matrices. Optional compute and save chunking to avoid GPU and CPU
  memory limits.

Options:
  -c, --chunks INTEGER       [default: 0]
  -s, --save-chunks INTEGER  [default: 0]
  --help                     Show this message and exit.
harrsha4 commented 1 month ago

Thank you for the reply. Another question, if possible: If you have a 3 cohort dataset that you want to include as a covariate, would we need to convert them to numeric and treat them as a numeric variable, or could we maintain a categorical variable?

liamgd commented 3 weeks ago

Currently, categorical covariates must be converted to numeric to be supported.