Recon8D utilizes oct-omics (genomics, histone PTMs, DNA methylation, transcriptomics, RNA splicing, miRNA, proteomics, and phosphoproteomics) to predict metabolomic variation across cancer lines from the Cancer Cell Line Encyclopedia, thereby inferring a multiomic prediction network of the metabolome.
File descriptions
RF_restuls: Pearson's correlations and p-values for all metabolite models from each of 8 omics classes.
example_datasets: metabolomics and histone PTM data that can be used to test the machine lerning code present in this repository.
omics_top_features: top 10 features for all metabolite models from each of 8 omics classes.
recon_mapping: MATLAB and Python scripts for extracting genes from reactions involving metabolites of interest and matching them with top feature lists.
ML_function.ipynb and ml_function.py: ML script for random forests, ridge regression, and lasso regression, along with example code for using histone PTM data as input.