IML

Integrative Machine Learning (IML)

Websites

Lichtarge Lab

Evolutionary Action

Lichtarge Lab Github

Related publications

Parvandeh et al., Consensus features nested cross-validation, Bioinformatics, 2020

Parvandeh et al., EPIMUTESTR: a nearest neighbor machine learning approach to predict cancer driver genes from the evolutionary action of coding variants, 2022

Katsonis P, Lichtarge O., A formal perturbation equation between genotype and phenotype determines the evolutionary action of protein coding variations on fitness, Genome Research, 2014 Sep 12. pii: gr.176214.114.

To run

1. Download the four data types from GDSC website

2. Annotate mutation data with Evolutionary Action

3. Preprocess four data types

   Rscript 1_Preprocessing_data.R

4. Predict IC50 values for pan-cancer cell lines

   Rscript 2_IML_PANCAN_1Dtype.R Rscript 3_IML_PANCAN_2Dtypes.R Rscript 4_IML_PANCAN_3Dtypes.R Rscript 5_IML_PANCAN_4Dtypes.R

5. Predict IC50 values for cancer specific cell lines

   Rscript 6_IML_CASpecific_1Datatype.R Rscript 7_IML_CASpecific_2Datatypes.R Rscript 8_IML_CASpecific_3Datatypes.R Rscript 9_IML_CASpecific_4Datatypes.R

6. Define sensitivity signature using genes that identified from step 5 for RNA-Seq

   Rscript 10_IML_SensitivitySignature.R

7. Visualize the results

   Rscript 11_Visualization.R

8. Validate on CCLE data

   Rscript 12_Validation_CCLE.R

Dependencies

install.packages(c('CORElearn', 'Rcpp', 'dplyr', 'parallel', 'foreach', 'doParallel', 'glmnet', 'randomForest', 'e1071', 'rpart'))

Other R packages

install.packages(c('ggplot2', 'tidyverse', 'hrbrthemes', 'viridis', 'ggpubr', 'ggbeeswarm', 'forcats', 'cvTools'))

Helper function (included)

Rcpp::sourceCpp("VCF2DM.cpp")