AvailabilityA new version of DL4DR based on the latest convolutional neural network (CNN) techniques will be available here, after the manuscripts are published. An old version (JavaDL) of deep learning program was developed based on Java for QSAR predictions, freely available for download here.
Implementation The easiest way to access our newest version of DL4DR is by pulling its Docker image from DockerHub ( Merkel, 2014) and running it using Docker engine. The image runs on Mac OS, Windows and various linux operating systems supported by the Docker engine. This takes care of all the dependencies as well as reproducibility issues. To run the image, users need to provide a folder path, SMILE strings in smi file format, and cell line name. The result files are generated in the provided folder. Additionally, users can run the image on HPC in parallel using either SLURM or SGE queue system without interfering with previously installed software or databases.
Commands 1) docker pull imdl/dl4dr 2) docker run -t -v folderpath:/home/dl4dr/output imdl/dl4dr output/[smifilename] [celllinename] Options 1) smifilename: Smi File Example 2) celllinename: At present, we provide models for the cell lines MDA-MB-231, HCC-1937 and MDA-MB-453. Example docker run -t -v C:/myfolder:/home/dl4dr/output imdl/dl4dr output/pyroles.smi MDA-MB-231