Preprocessing

3. Preprocessing: Standardasation and scaling of gene expression data

Prerequisite: Data files must have been downloaded and initial classification applied as described in DataLoading.

The Jupyter Notebook in this folder will prepare the RNA sequencing data with the following process:

• Load the file generated in the previous step (clinical.csv) as a Pandas dataframe;
• Filter to include only genes with the biotype protein_coding;
• Extract expression values from the stranded_first column;
• Format the data into a single row, with gene names as columns;
• Add clinical metadata including TNBC label and Case ID;
• Handle duplicate case IDs and gene colums;
• Save the resulting dataframe to a new file geneDataPreProcessed.csv in the Data folder (this will not be included in the repository);

The resulting preprocessed data file can be loaded with the following code:

dataPath = '../Data'
df = pd.read_csv(os.path.join(dataPath, 'geneDataPreProcessed.csv'))

Next step: Jupyter Notebooks that parse the preprocessed data, select features using different methods, and link them from all RNA Seq files, can be found in the Features folder that lives next to this Preprocessing folder.

Key findings

Each file contained expression data for 19,354 protein-coding genes. 24 duplicate genes have been handled. No duplicate case IDs were found.