Skip to content

README.md

Latest commit

 

History

History
79 lines (59 loc) · 4.29 KB

README.md

File metadata and controls

79 lines (59 loc) · 4.29 KB

differential_expression_analysis

This project is based on differential expression analysis with the following characteristics:

  • It uses the lung adenocarcinoma (LUAD) dataset from The Cancer Genome Atlas Program (TCGA), obtained from cBioportal.
  • It uses the SEX variable to compare genes and obtain the differentially expressed ones.
  • Use the Limma library of R to perform the DEA.

Requerimientos

  • R version 4.4.2

Descargar datasets

Ejecutar el siguiente comando para obtener todos los datasets de cbioportal que tienen datos de RNA-Seq:

Rscript download_rnaseq_datasets.r

Run DEA

Use the following command to perform differential expression analysis

Rscript aed_limma_cbioportal_TMP.r

The script downloads all necessary libraries before running the analysis.

Operating details

Of the dataset downloaded from cBioportal, only the files "datasets/luad_tcga/data_clinical_patient.txt", "datasets/luad_tcga/data_clinical_sample.txt", and "datasets/luad_tcga/data_mrna_seq_v2_rsem.txt" are used for the analysis. These correspond to the patient clinical data, the sample data, and the gene expression data, respectively.

Expression data: The dataset has RNASeq data expressed in Transcripts per million (TPM) values ​​using RSEM.

Processing of clinical and sample data

  • Filter only the PATIENT_ID and SEX columns in clinical data.
  • Convert SEX values ​​to uppercase in clinical data.
  • Remove duplicate records in clinical data.
  • Filter only the SAMPLE_ID and PATIENT_ID columns in sample data.
  • Perform a merge between sample_data and clinical_data using PATIENT_ID as the key.
  • Keep only the SAMPLE_ID and SEX columns.
  • Replace hyphens (-) with periods (.) in SAMPLE_ID.
  • Ensure there are no duplicates by removing them from the dataset (there should be no duplicates at this point).
  • Count the number of samples per sex, and calculate the percentage of each sex in the sample.
  • Store the entire processed dataset in a dataframe called 'metadata'

RNASeq dataset processing

  • Genes with duplicate names (Hugo_Symbol) are identified.
  • Duplicate genes are removed, keeping only the first occurrence of each Hugo_Symbol.
  • The 'Entrez_Gene_Id' column is removed.
  • Rows where Hugo_Symbol is NA are removed.
  • Hugo_Symbol is assigned as row names.
  • Samples in metadata that are not present in rnaseq_tpm_data are removed.
  • The result is returned in a data frame named 'rnaseq_tpm_data'.

DEA with limma

  • Create boxplots to compare expression distributions between samples.
  • Convert TPM values ​​to a log2 scale (TPM + 1) to stabilize variance.
  • Generate histograms of the expression distribution before and after the log2 transformation.
  • Save the plots to a plots.pdf file.
  • Create a design matrix with the variable SEX as a factor.
  • Calculate the variance of each gene in the log-transformed data.
  • Eliminate genes with a variance less than the threshold (1e-4 by default).
  • Calculate the mean expression of each gene, set a percentile-based threshold (15% by default), and then eliminate genes with mean expression below this threshold.
  • Use the limma package to identify genes differentially expressed by sex (SEX), using the dataset modified with the two points above.
  • Fit a linear model (lmFit) and apply the eBayes adjustment.
  • Extracts the results with corrected p-values ​​(Benjamini-Hochberg correction).
  • Returns the table of differentially expressed genes.
  • Sorts the results of the differential analysis by the adjusted p-value (adj.P.Val), in ascending order (lowest to highest).
  • Sorts the results by the fold change (logFC), in descending order (highest to lowest).
  • Sorts the results in two steps: first by the adjusted p-value in ascending order and then by logFC in descending order.
  • Extracts the 50 genes with the lowest adjusted p-values ​​(the most significant).
  • Generates a "Volcano Plot" to visualize the relationship between logFC and -log10 (adjusted p-value). Distinguishes significant genes (in red) from non-significant ones (in black).
  • Saves the results in two CSV files (one with all results and one with the top 50)