Currently it uses Wanderlust to calculate developmental trajectories within a population of single cell data. The input is a matrix of single-cell measurements for multiple features, and the output is a value for each cell representing its developmental stage within the population evolution timeline.
Cells more similar to the initial cell (seed) will have a lower score than cells that have matured for longer time than the seed. It is useful to try several seed cells, or use existing knowledge of stemness-like markers. If the seed cell is in a mature stage it will still predict the further stages.
When using this component, please cite Wanderlust: http://www.c2b2.columbia.edu/danapeerlab/html/wanderlust.html The code in folder wanderlust is original from the CYT package.
See paper for details of the parameters. In brief; first generates num_graphs klNN graphs, then, using seed as starting point, choose num_landmarks random landmarks. Continues by weighting landmark-node pairs by distance, metric, to the power of k and adjusting each cell's position until convergence.
Recommended parameters to tweak are: num_landmarks, metric, num_graphs, k, and l.
Version | 1.0 |
---|---|
Bundle | tools |
Categories | Analysis |
Authors | Julia Casado (julia.casado@helsinki.fi) |
Issue tracker | View/Report issues |
Requires | Matlab |
Source files | component.xml timeline.m |
Usage | Example with default values |
Name | Type | Mandatory | Description |
---|---|---|---|
in | CSV | Mandatory | High-dimensional single-cell data table. Columns represent features or markers, and rows represent cells or datapoints. First column contains Event or Cell ID, and first row contains the column names |
Name | Type | Description |
---|---|---|
out | CSV | One column with the stage for each data point, the stage is part of a continuum of the hypothetical distance between the most immature and the most mature cell in the population. |
Name | Type | Default | Description |
---|---|---|---|
band_sample | boolean | false | Whether the landmarks are equidistant or randomly distributed. |
branch | boolean | false | Whether wanderlust will seek for a branch. |
k | int | 5 | Truncate the neighbors to random k out of l. |
l | int | 30 | Size of neighborhood l>k closest points. |
metric | string | "euclidean" | Distance metric for constructing the nearest neighbor graphs. |
num_graphs | int | 20 | Number of random selections of k out of l, that is, number of repetitions. |
num_landmarks | int | 20 | Number of waypoints\landmarks. |
partial_order | string | "" | A list of integers that represent datapoints in data. If non-empty, the indices point to landmarks and their order is forced. |
seed | int | -1 | Index of starting cell. If -1 the algorithm will choose a random cell. |
snn | int | 1 | Number of shared nearest neighbors. |
voting_scheme | string | "exponential" | How to weigh each point's contribution. Valid values: uniform, exponential, linear, inverse-linear. |
Test case | Parameters▼ | IN in |
OUT out |
|||
---|---|---|---|---|---|---|
case1 | properties | in | (missing) | |||
num_graphs = 10, |