4DMax is a tool for prediction of the 4dgenome conformation from a time-series Hi-C dataset

Install

4DMax runs entirely in python. Use expects access to a GPU. To install use 'conda env create -f environment.yml'

4DMax

1. Format your time series Hi-C data
Each Hi-C experiment should be represented as a 3 column tab seperated text file where (pos1, pos2, val)
2. Build dataset and hyper parameters configuration files
examples shown in:

1. Config/Datasets/example.json

2. Config/Hyper_Params/example.json

   Hyper Params

   • eta: weight of movement loss
   • alpha: contact map to distance constraint conversion ratio IF=d^alpha
   • lr: learning rate
   • epoch: number of epochs to train

   Data Set

   • name: genomic Process name
   • step: granularity of 4D Model
   • chro: chromosome number
   • rep: biological replicate number
   • taos: indx of hi-c experiments in time process
   • datasets: hi-c experiment text files

3. Run 4DMax 'python 4dmax.py {input.dataset} {input.param}'

4 View Strucutes 'python Python_Scripts/create_gif.py {output.outfig} {input.npfile}

Reproduce white paper experiments.

1. Download needed Hi-C files Cardiomyocyte GSE106690 iPSC GSE96611
2. Generate modes 'snakemake'
3. Generate TADS 'cd TADS; snakemake --use-conda '
4. Generate AB Compartments: 'cd AB; snakemake'