Does this skill support single-cell ATAC-seq?

Yes, the scEmbed feature is specifically designed for single-cell ATAC-seq analysis, including cell-level embeddings and integration with Scanpy workflows.

How do I train genomic embeddings with this skill?

You can use the Region2Vec capability to tokenize BED files against a universe reference and train unsupervised word2vec-style embeddings for genomic regions.

What is Geniml used for?

Geniml is a Python library for applying machine learning to genomic interval data (BED files), offering tools for embeddings, single-cell analysis, and universe building.

What are consensus peaks or 'universes' in Geniml?

Universes are reference peak sets created from collections of BED files using statistical methods like coverage cutoffs or Hidden Markov Models to standardize genomic regions.

Is Geniml compatible with Scanpy?

Yes, Geniml integrates seamlessly with Scanpy; scEmbed embeddings can be stored directly as obsm entries in AnnData objects for clustering and UMAP visualization.

Genomic Interval Machine Learning (Geniml)

Name: Genomic Interval Machine Learning (Geniml)
Author: Zehong-Wang

byZehong-Wang

0•

Data Science & ML

Builds and deploys machine learning models for genomic interval data and single-cell chromatin accessibility analysis.

Geniml is a comprehensive Python toolkit designed for unsupervised machine learning on genomic regions stored in BED files. It enables researchers to generate high-dimensional embeddings for genomic intervals using Region2Vec, integrate metadata with region sets via BEDspace, and perform advanced single-cell ATAC-seq analysis with scEmbed. By providing standardized workflows for tokenization, universe building, and embedding evaluation, Geniml bridges the gap between raw genomic data and sophisticated downstream ML tasks like clustering, similarity searching, and cell-type annotation.

Key Features

01Genomic data utilities for caching, randomization, and evaluation

02Statistical universe building for consensus peak sets

03scEmbed for single-cell ATAC-seq dimensionality reduction

040 GitHub stars

05BEDspace for joint region and metadata embedding models

06Region2Vec for unsupervised genomic region embeddings

Use Cases

01Clustering and visualizing single-cell ATAC-seq datasets

02Training embedding models to represent genomic features as vectors

03Building searchable genomic databases with cross-modal metadata queries

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add zehong-wang/kosmos geniml

For use in Claude.ai and ChatGPT

Download Skill