unsupervised_analysis
unsupervised_analysis copied to clipboard
epigen
A general purpose Snakemake workflow and MrBiomics module to perform unsupervised analyses (dimensionality reduction & cluster analysis) and visualizations of high-dimensional data.
**Summary:** Clustering stability could be assessed by doing multiple clusterings, always randomly sampling 90% of the data. Consensus approach could be used to extract stable clustering. **Drawbacks:** Computationally expensive **Open...
for details see @sreichl [Master Thesis](https://catalogplus.tuwien.at/primo-explore/fulldisplay?docid=UTW_alma2180892620003336&context=L&adaptor=Local%20Search%20Engine&vid=UTW&lang=de_DE&search_scope=UTW&tab=default_tab&query=addsrcrid,exact,AC15010381) Chapter 4.3.2. The Consensus Approach
features_to_plot option "all” plots all features with a warning that this is only useful with relatively low dimensional data. provide a number eg
- Diagnostics - Dynamic visualization of high-dimensional data - https://www.nature.com/articles/s43588-022-00380-4 - https://github.com/sunericd/dynamicviz
- https://github.com/rongstat/meta-visualization - https://www.nature.com/articles/s41467-023-36492-2 - which parametrization to use? Maybe a two-step thing in the user manual, first find one representation/parameter set per dataset and method and then apply meta-vis.
graph based alternatives - Infomap, but literature claims: Leiden>Louvain>Infomap → Leiden - Spectral clustering - https://towardsdatascience.com/spectral-clustering-aba2640c0d5b - https://en.wikipedia.org/wiki/Spectral_clustering
- on densMAP results i.e., embedding - https://github.com/scikit-learn-contrib/hdbscan - config lambda (as list?)