Does this skill support single-cell ATAC-seq?

Yes, Geniml includes the scEmbed module specifically for single-cell ATAC-seq, enabling cell-level embeddings and seamless integration with the scanpy ecosystem.

What machine learning models does Geniml support?

Geniml supports several specialized models including Region2Vec (word2vec for regions), BEDspace (StarSpace-based joint embeddings), and various statistical models for universe building like HMM and Maximum Likelihood.

What is the primary purpose of Geniml?

Geniml is designed to apply machine learning to genomic interval data, allowing users to create unsupervised embeddings of regions (BED files) for clustering, search, and downstream analysis.

What are 'universes' in the context of Geniml?

Universes are consensus peak sets created from multiple BED files. They serve as a standardized reference for tokenizing genomic regions before training machine learning models.

Geniml Genomic Machine Learning

Name: Geniml Genomic Machine Learning
Author: henriquescastilho

byhenriquescastilho

•

Data Science & ML

Builds machine learning models and unsupervised embeddings for genomic interval data and single-cell chromatin accessibility datasets.

Geniml is a powerful toolkit for computational biologists and data scientists working with genomic interval data (BED files). It provides specialized methods for learning high-dimensional embeddings of genomic regions, single cells, and metadata, enabling complex tasks like similarity searches, clustering, and cross-modal queries. Whether you are building consensus peak universes, training Region2Vec models for bulk data, or analyzing single-cell ATAC-seq with scEmbed, Geniml offers the statistical rigor and machine learning architecture necessary to transform raw genomic coordinates into actionable biological features.

Key Features

01Specialized single-cell ATAC-seq cell-level embeddings and scanpy integration

02Comprehensive utility suite for BED file randomization, caching, and evaluation

03Joint region and metadata embedding for cross-modal similarity search with BEDspace

04Statistical consensus peak (universe) building using CC, ML, and HMM methods

051 GitHub stars

06Unsupervised learning of genomic region embeddings using Region2Vec

Use Cases

01Cell-type annotation and clustering for single-cell chromatin accessibility analysis

02Building searchable genomic databases that link experimental metadata to specific region sets

03Dimensionality reduction and feature engineering for large-scale BED file collections

Key Features

01Specialized single-cell ATAC-seq cell-level embeddings and scanpy integration

02Comprehensive utility suite for BED file randomization, caching, and evaluation

03Joint region and metadata embedding for cross-modal similarity search with BEDspace

04Statistical consensus peak (universe) building using CC, ML, and HMM methods

051 GitHub stars

06Unsupervised learning of genomic region embeddings using Region2Vec

Use Cases

01Cell-type annotation and clustering for single-cell chromatin accessibility analysis

02Building searchable genomic databases that link experimental metadata to specific region sets

03Dimensionality reduction and feature engineering for large-scale BED file collections