# scanpy.external.tl.harmony_timeseries¶

scanpy.external.tl.harmony_timeseries(adata, tp, n_neighbors=30, n_components=1000, n_jobs=- 2, copy=False)

Harmony time series for data visualization with augmented affinity matrix at discrete time points [Nowotschin18i].

Harmony time series is a framework for data visualization, trajectory detection and interpretation for scRNA-seq data measured at discrete time points. Harmony constructs an augmented affinity matrix by augmenting the kNN graph affinity matrix with mutually nearest neighbors between successive time points. This augmented affinity matrix forms the basis for generated a force directed layout for visualization and also serves as input for computing the diffusion operator which can be used for trajectory detection using Palantir.

Note

Parameters
adata : AnnDataAnnData

Annotated data matrix of shape n_obs × n_vars. Rows correspond to cells and columns to genes. Rows represent two or more time points, where replicates of the same time point are consecutive in order.

tp : strstr

key name of observation annotation .obs representing time points. Time points should be categorical of dtype=category. The unique categories for the categorical will be used as the time points to construct the timepoint connections.

n_neighbors : intint (default: 30)

Number of nearest neighbors for graph construction.

n_components : (default: 1000)

Minimum number of principal components to use. Specify None to use pre-computed components. The higher the value the better to capture 85% of the variance.

n_jobs : intint (default: -2)

Nearest Neighbors will be computed in parallel using n_jobs.

copy : boolbool (default: False)

Return a copy instead of writing to adata.

Return type
Returns

Depending on copy, returns or updates .obsm, .obsp and .uns with the following:

X_harmony - ndarray (obsm, dtype float)

force directed layout

harmony_aff - spmatrix (obsp, dtype float)

affinity matrix

harmony_aff_aug - spmatrix (obsp, dtype float)

augmented affinity matrix

harmony_timepoint_var - str (uns)

The name of the variable passed as tp

harmony_timepoint_connections - ndarray (uns, dtype str)

Example

>>> from itertools import product
>>> import pandas as pd
>>> from anndata import AnnData
>>> import scanpy as sc
>>> import scanpy.external as sce


Load AnnData

A sample with real data is available here.

Random data sets of three time points with two replicates each:

>>> adata_ref = sc.datasets.pbmc3k()
>>> start = [596, 615, 1682, 1663, 1409, 1432]
...     *(adata_ref[i : i + 1000] for i in start),
...     join="outer",
...     batch_key="sample",
...     batch_categories=[f"sa{i}_Rep{j}" for i, j in product((1, 2, 3), (1, 2))],
... )
....    time_points, categories=['sa1', 'sa2', 'sa3']
... )


Normalize and filter for highly expressed genes

>>> sc.pp.normalize_total(adata, target_sum=10000)

>>> sce.tl.harmony_timeseries(adata, tp="time_points", n_components=500)

>>> sce.pl.harmony_timeseries(adata)