vocalLearning_TACIT

Includes script and commands for generating datasets for putamen rat-only model, and putamen multi-species model with TACIT pipeline.
Followed with permulation methods for three different physical traits.

• multiSpeciesModelPrediction
  Using trainied putamen multi-species model to make predictions on OCRs of 224 mammals.

• negativeGCMatched
  Generates random GC matched dataset as negatives using BiasAway.

• putamenMultiSpeciesModelDataset
  Generates training, validation, evaluation used in multi-species (rat, macaque, bat) putamen model.

• phylogeneticPermulation
  Find association between predicted enhancer activities and physical traits, including longevity(Maximum life span), total daily sleep, and vocal learning, with TACIT.

• ratModelDataset
  Generates training, validation, evaluation used in rat putamen model.

• ratPutamenSVM: uses SVM to provide insights for CNN hyperparameter tuning.

• humanPutamenDataset: Generates training, validation, evaluation used in human putamen model.

Automated scripts

• TACIT pipelines: automate dataprocessing including filtering, plotting, and generating matrix with permulationList.py, plotPDist.R, bhCorrection.R, qValue.R, filterNFasta.py.
• ATAC seq helpers: automate mapping peaks from halLiftover and HALPER with halperMapPeak.sh. Ortholog conversion for bat sequences with batFormatConvert.sh.
• Model Prediction interpretation: automate testing p value distribution shift comparing prediction and literature with pValShiftTest.sh and plotWilcoxon.py

Dependencies

• Conda environment: keras2-tf27.yml used for CNN training/predictions. For other scripts, use hal.yml.
• Most scripts require: filterPeakName.py from OCROrthologPrediction; convertChromNames.py from TACIT.
• Phyloenetic permulation, requires TACIT.