If you prefer to use pip to install DeepDETAILS, you may have to install some non-Python dependencies yourself.

For bedGraphToBigwig (and other standalone programs made by UCSC Genome Browser team): you can directly download the binary version from the UCSC FTP site

1. Download the tools from: UCSC Genome Browser Utilities
2. Choose the version appropriate for your operating system (Linux, Mac, or Windows). For Linux, you can download it from the directory: http://hgdownload.soe.ucsc.edu/admin/exe/linux.x86_64/
3. Once downloaded, you can add the location of the binary (e.g., bedGraphToBigwig) to your system's PATH, or you can simply move it to a directory that's already in your PATH (e.g., /usr/local/bin/).
4. After that, you should be able to run the command like this: bedGraphToBigwig input.bedGraph genome.size output.bigWig

For bedtools, it's available from many package managers. If you are an Ubuntu user, you can use sudo apt-get install bedtools.

If you encounter with errors like the following

torch.cuda.OutOfMemoryError: CUDA out of memory. Tried to allocate 160.00 MiB (GPU 0; 23.69 GiB total capacity; 22.24 GiB already allocated; 114.06 MiB free; 23.19 GiB reserved in total by PyTorch) If reserved memory is >> allocated memory try setting max_split_size_mb to avoid fragmentation.  See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CONF

It means you don't have enough GPU memory to support the deconvolution of your sample with the current DeepDETAILS configuration. Here are several things you can try:

• Use the fused version of DeepDETAILS instead of the sequence only model (don't add --seq in your command)
• Use a smaller batch size, the default batch size is 32, smaller values such as 16 or 8 can be helpful. Example: --batch-size 16
• Aggregate similar cell types to reduce the total number of deconvolution target