-
Added abilitiy to import/export expression data from/to Scanpy (see new tutorial!).
ExpMatrix.from_anndata()
imports anAnnData
object (Scanpy)- Conversely,
matrix.to_anndata()
exports the expression matrix as anAnnData
object.
-
Added utility function
util.zscore()
to convert a matrix to z-scores.
-
Added support for heatmaps (
workflows.overexpressed_gene_heatmap()
) -
Improved data preprocessing (
preprocess.preprocess_data()
) that allows better reporting of QC metrics -
Support for graph-based clustering using the Leiden algorithm, by running Scanpy's implementation using the Monet latent space model (
workflows.graph_based_clustering()
) -
Latent space model simplified => Now everything is built around the
PCAModel
class -
Nearest-neighbor aggregation function optimized (now uses scikit-learn
NearestNeighbor
class) -
Support for processing data produced with STARsolo's Velocyto option
-
New concepts of workflows that provide a simplified API to streamline common analysis tasks (try
from monet import workflows as flow
), including:- Data preprocessing (e.g.,
workflows.preprocess_10x_v3()
) - Visualization (e.g.,
workflows.tsne()
) - Clustering (e.g.,
workflows.graph_based_clustering()
)
- Data preprocessing (e.g.,