scNoah

Benchmarking cell type annotation and modality prediction.

Balance dataset

Balancing dataset using your own annotation for future model training. Oversmaple or undersample some cell types.

scnoah.balance	Balance cell types by undersampling large classes and oversampling small classes with PCA-based synthetic-cell interpolation.
scnoah.oversample	Oversample underrepresented cell types in an AnnData object using PCA-based interpolation.
scnoah.undersample	Undersample overrepresented cell types in an AnnData object.

Annotation metrics

Test annotation method quality using confusion matrix, accuracy, balanced accuracy and calculating cell type specific precision, recall (also called sensitivity), specificity, f1-score, geometric mean, and index balanced accuracy of the geometric mean.

scnoah.report_classif_full	Returns metrics (precision, recall (also called sensitivity), specificity, f1-score, geometric mean, and index balanced accuracy of the geometric mean) of predicted cell types.
scnoah.report_classif_sens_spec	Returns specificity and recall (also called sensitivity) metrics of predicted cell types.
scnoah.conf_matrix	Compute confusion matrix to evaluate the accuracy of a classification.
scnoah.pred_status	Find correct and incorrect predictions.

Regression metrics

Test modality prediction method quality using error metrics (RMSE, MedianAE, MeanAE), EVS, R² score, PCC, SCC and KTCC. Also, visualise metrics on cell embeddings.

RMSE - Root mean squared error, MeanAE - Mean absolute error, MedianAE - Median absolute error, EVS - Explained variance score, R² score - Coefficient of determination, PCC - Pearson correlation coefficient, SCC - Spearman correlation coefficient, KTCC - Kendall tau correlation coefficient.

For error metrics (RMSE, MedianAE, MeanAE): lower value - better prediction

scnoah.report_reg	Returns multiple metrics of cell surface proteins prediction.
scnoah.regres_status	Compute regression status of cells to visualize on UMAP.
scnoah.pearson_coef	Compute Pearson correlation coefficient of predicted feature.
scnoah.spearman_coef	Compute Spearman correlation coefficient of predicted feature.
scnoah.kendalltau_coef	Compute Kendall’s tau, a correlation measure of predicted feature.

Count cells

Count number of cell types per sample or condition.

scnoah.cell_counter

Count cell types in samples.

Explanations

Get explanations of gene importances in scAdam model prediction.

scnoah.explain	Compute SHAP explanations for the predicted probability of one cell type.
scnoah.feature_importance_df	Create a feature-importance table from SHAP explanations.

Managing large datasets

Get fraction of large dataset.

scnoah.get_frac	Return a random subset of an AnnData object.
scnoah.get_samples	Get samples from AnnData object.

Difference between clusters

Calculate Integral of absolute density difference and Mutual Information between two clusterings.

scnoah.clust_diff

Calculates metrics to Integral of absolute density difference and Mutual Information between two clusterings.