Geniml

If you're doing machine learning on genomic intervals, this handles the weird stuff between BED files and embeddings. Region2Vec gives you unsupervised region embeddings, BEDspace adds metadata-aware search, and scEmbed does single-cell ATAC-seq clustering that drops into scanpy workflows. The universe building tools are honestly the MVP here since tokenization quality makes or breaks everything downstream. It's not trying to be a general genomics Swiss army knife, just focused on turning regions into vectors you can actually work with. Assumes you know your way around BED files and have PyTorch installed if you want the ML features. The CLI is there but most real work happens in Python.