scATAC数据分析哪家强

paper

workflow

article: 10.1186/s13059-024-03356-x, Genome Biology, 2024

challenges

  1. sparse and noisy signal <- low copy number

  2. no fixed feature sets

Packages and algorithms

  1. Signac: LSI + SVD

  2. ArchR: LSI

  3. cisTopic: LDA

  4. SnapATAC: diffusion map

  5. SnapATAC2: Laplacian epigenmaps

  6. BROCKMAN: gap k-mer frequency

datasets

6 published datasets of divergent sizes and sequencing protocols and from different tissues and species

Assesment

cell embeddings, graph structure, and final partitions <- 10 metrics

Results

  1. Feature aggregation: SnapATAC & SnapATAC2 > LSI-based methods.

  2. For datasets with complex cell-type structures, SnapATAC and SnapATAC2 were the most effective.

  3. For large datasets, SnapATAC2 and ArchR were the most scalable in terms of computational efficiency.

最后编辑于 ：
©著作权归作者所有,转载或内容合作请联系作者
平台声明：文章内容（如有图片或视频亦包括在内）由作者上传并发布，文章内容仅代表作者本人观点，简书系信息发布平台，仅提供信息存储服务。

推荐阅读更多精彩内容