Analysis tools for UNCURL

Installation

1. Clone the repository.
2. Run pip install -r requirements.txt
3. In the main directory, run pip install .

Cluster-specific gene expression

These methods help to identify the genes that are overexpressed in each cluster.

The gene_extraction.pairwise_t function returns two 3d arrays of shape (k, k, genes),

The outputs, c_scores and c_pvals, are dicts of cluster labels to tuples (gene_index, c_score_or_pval)

Example:

from uncurl_analysis import gene_extraction

scores, pvals = gene_extraction.pairwise_t(data, labels)
c_scores, c_pvals = gene_extraction.c_scores_from_t(scores, pvals)

Identifying similar bulk datasets

Given some known datasets and a single-cell query dataset, we can get a similarity score between the query dataset and each of the known datasets.

This assumes that all genes are already aligned between the two datasets.

Example:

from uncurl_analysis import bulk_data

# bulk_means is a dict of label : array. cell is a 1d array.
scores = bulk_data.bulk_lookup(bulk_means, cell, method='poisson')

SCAnalysis: a framework for managing scRNA-seqdata analysis