What type of data does Geniml support?

Geniml primarily works with genomic interval data in BED format, including ChIP-seq and ATAC-seq datasets.

What is a 'Universe' in the context of this skill?

A Universe is a reference set of consensus peaks used to standardize and tokenize genomic regions across multiple datasets.

Can I use Geniml for single-cell analysis?

Yes, the scEmbed capability is specifically designed for single-cell chromatin accessibility embeddings and integration with scanpy workflows.

Does Geniml support joint metadata searches?

Yes, the BEDspace feature allows you to create a shared embedding space for both genomic regions and their associated experimental metadata labels.

Genomic Machine Learning (Geniml)

Name: Genomic Machine Learning (Geniml)
Author: pur3v4d3r

bypur3v4d3r

•

데이터 과학 및 ML

Builds machine learning models and unsupervised embeddings for genomic interval data and BED file collections.

소개

Geniml is a specialized toolkit designed for processing genomic interval data from BED files into machine-learning-ready formats. It enables the creation of unsupervised region embeddings using Region2Vec, joint metadata-region analysis via BEDspace, and single-cell chromatin accessibility studies with scEmbed. By standardizing the way genomic 'universes' are built and tokenized, Geniml allows researchers to perform similarity searches, clustering, and downstream analysis on large-scale chromatin accessibility datasets with statistical rigor.

주요 기능

1 GitHub stars
Region2Vec for unsupervised genomic region embeddings
Automated 'Universe' building for consensus peak sets
scEmbed for single-cell ATAC-seq dimensionality reduction
BEDspace for joint embedding of regions and metadata labels
BED file tokenization and evaluation utilities

사용 사례

Clustering and cell-type annotation for scATAC-seq data
Building feature vectors for supervised genomic ML models
Similarity searches across large BED file repositories

소개

주요 기능

1 GitHub stars
Region2Vec for unsupervised genomic region embeddings
Automated 'Universe' building for consensus peak sets
scEmbed for single-cell ATAC-seq dimensionality reduction
BEDspace for joint embedding of regions and metadata labels
BED file tokenization and evaluation utilities

사용 사례

Clustering and cell-type annotation for scATAC-seq data
Building feature vectors for supervised genomic ML models
Similarity searches across large BED file repositories