Machine_learning

Machine learning Project - mRNA expression data

Goal

Test different classification methods (discuss) → Select the best method

Dataset

Gene expression (Base ML dataset, Raw RNA seq dataset) of five different cancer types.

Methods

Preprocessing (Unsupervised learning):

• Principal Component Analysis(PCA)

• tSNE

Use different classification methods (Supervised learning):

• K-Nearest Neighbors

• Linear Models

• Naive Bayes Classifiers

• Decision Trees

• Kernelized Support Vector Machines(?)

TODO

Keep track on what we still have to do. Please update this list with new todo's.

• Update README.
• Investigate preprocessing that is applied to the data.
• Write about preprocessing steps in report.
• Keep track on references in the report.
• Reorganize repository (give logical filenames, restructure folders, etc.).
• Rewrite PCA scripts structure.
• Calculate amount of PC's needed (PCA script).
• Review PCA script (especially investigate explained variation values).
• Download data of different cancer types from Synapse and merge with annotations (also from Synapse).
• Try different Hyperparameters in the ML algorithms (Knn, SVM, ecc) and cross validation
• PCA: try to apply it within cancer types
• find important features → DEG (Differentially expressed genes)
• KEGG analysis (Pathways)
• check for class imbalances (bar plot)