Data Preparation Modules: panoply_cosmo

panoply_cosmo

This is a PANOPLY implenentation of the COSMO toolset. Full COSMO documentation here.

Inputs

Mandatory inputs:

• STANDALONE (Boolean): Determines whether COSMO should run as part of panoply_main pipeline with inputs from panoply_harmonize, or if COSMO should run independently with user-selected .gct inputs.
• yaml_file (File): Parameters yaml file from PANOPLY setup. Must contain gene.id.col and cosmo default parameters in the cosmo.params section.
• label (String): Label to use for report filename.

Mandatory inputs if STANDALONE == "false":

• panoply_harmonize_tar (File): Tar output from panoply_harmonize module. The inputs to the cosmo functions are taken from the harmonized-data folder (proteome matrix, RNA matrix, sample annotations).
• ome_type (String): The PANOPLY omics type of the input (e.g. "proteome").

Mandatory inputs if STANDALONE == "true":

• d1_file (File): A .gct file. Typically proteome or other protein-level data.
• d2_file (File): A .gct file. Typically RNAseq file to be compared with d1_file for sample mislabeling.
• sample_file (File): A .csv file with sample annotations.

Optional inputs:

• sample_label (String): Column header(s) in sample_file to use for prediction. This typically includes gender information. Column will be excluded if it is unbalanced (likely to cause an error). If multiple inputs, separate inputs with a comma (e.g. gender,msi). If no input, columns are pulled from default set in yaml file.
  • Note: if you specifically use gender as a sample label, then COSMO will know that column contains male/female information and will use genes from sex chromosome to better predict this attribute. COSMO does not properly identify other synonymous column titles (such as sex or Gender).
• run_cosmo (Boolean): whether or not to actually run the cosmo functions. If false, the original tar file is saved as the final output and the cosmo html report is not generated. If no input, run_cosmo is pulled from default set in yaml file.

Output

• cosmo_tar (File): The main tar output. When STANDALONE == "false", this tar has the same data and format as the original tar input (plus cosmo results if cosmo was run).
• cosmo_report_html (File): The html report summarizing cosmo results.

Common Pitfalls

The main source of computing errors in COSMO is improper selection of sample labels. PANOPLY-specific preprocessing should eliminate sample labels that are likely to cause errors. COSMO requires clinical attributes that are well-balanced with only two levels (e.g. male/female, positive/negative) and no NA's. These are used to predict if there is mislabeling between the sample annotation file (sample_file) and any of the data files (d1_file, d2_file).

If you get the following error message, this likely means that one of the sample label columns in not well-balanced. This happens because COSMO divides up the samples into 5 cross-validation sets, and if any of those sets contains a class with 1 or 0 observations then COSMO cannot build the GLM model. Consider choosing an alternative sample label.

<simpleError in { which = foldid == i if (length(dim(y)) > 1) y_sub = y[!which, ] else y_sub = y[!which] if (is.offset) offset_sub = as.matrix(offset)[!which, ] else offset_sub = NULL glmnet(x[!which, , drop = FALSE], y_sub, lambda = lambda, offset = offset_sub, weights = weights[!which], ...)}: task 4 failed - "one multinomial or binomial class has 1 or 0 observations; not allowed">