ComBind

Prerequisites

R software version > 3.5.2

Required R-packages: ShortRead, randomForest, reshape2, pROC

Ready trained random forests

Trained with 75% of CAP-SELEX sequences, negative set constructed by sequence shuffling

https://drive.google.com/drive/folders/1r78j_PLeG91CSYB0F3B3ildL5s1odJ1M

Train ComBind

Place training set DNA

Place positive training sequences to: DNA/CAP-SELEX-Positive-Training/

Place negative training sequences to: DNA/CAP-SELEX-Negative-Training/

1. Name the data sets in training folders with the same name
2. Ideally, positive and negative data sets in training should be balanced in size
3. Format: fasta.gz
4. Sequence length is at least 25 nucleic acids, preferably 40 nucleic acids, while upper limit is not defined
5. All sequences must have the same length
6. Sequences with ambigious letters/nucleotides are removed from random forest training

Train ComBind

Run from the command line in ComBind folder:

Rscript ./CODE/Train_ComBind.R DATASET_NAMES_IN_DNA_FOLDER TF1_NAME TF2_NAME

NOTE: Make sure that 'PWM_HT-SELEX' folder include TF1_NAME and TF2_NAME PWMs

Example for Alx4-Tbx21 using a subset of CAP-SELEX DNA sequences:

Rscript ./CODE/Train_ComBind.R ALX4_TBX21_SUBSET ALX4 TBX21

Test ComBind

Place test set DNA

Place sequences you wish to test with ComBind to: DNA/SCORE-DNA/

1. Format: fasta.gz
2. Sequence length is at least 25 nucleic acids, preferably 40 nucleic acids, while upper limit is not defined (sequence length can be different from training sequence length)
3. All test sequences must have the same length
4. Test sequences should not entail ambigious letters (only "A", "T", "C" and "G" are allowed)

Test ComBind

Run from the command line in ComBind folder:

Rscript ./CODE/Test_ComBind.R TESTSET_NAME_IN_DNA_FOLDER RF_NAMES_IN_RF_FOLDER_WITHOUT_RF_NUMBER_PREFIXES

Example for Alx4-Tbx21 using a subset of CAP-SELEX DNA sequences:

Rscript ./CODE/Test_ComBind.R ALX4_TBX21 ALX4_TBX21_SUBSET