RNASeq data classifier
Background: Kidney transplant recipients with “operational tolerance” (OT) maintain a functioning graft without immunosuppressive (IS) drugs, thus avoiding treatment complications.
Problems:
- IS drugs can influence gene-expression signatures aiming to identify OT among treated KTRs.
- Small number of samples instead of enormous number of genes.
What have I done for now:
- Simple classifier based on Logistic Regression.
- For a more accurate classification, get rid of "unnecessary" genes by using feature selection, which was performed within the 20 x 5-fold cross validation.
In future:
- generate artificial data
- maybe try some other classifiers (SVM, Adaboost)
- confounding control...