ALPODs Operator

Description

The ALPODs (Algorithmic Population Descriptions) Operator performs cluster classification on cytometry data. It builds a decision tree to identify and classify cell populations based on marker expressions.

Usage

Input projection	.
row	represents the variables (e.g., markers, channels)
column	represents the observations (e.g., cells, events)
y-axis	measurement values

Input parameters	.
verbose	boolean, enables detailed logging, default is true
min_samples	numeric, minimum number of samples required to split a node, default is 10
purity_threshold	numeric, purity threshold for stopping node splitting, default is 0.95
effect_size_threshold	numeric, effect size threshold for relevant subpopulations, default is 0.5

Output relations	.
cell_type	character, the identified cell type/cluster
rule	character, the classification rule used

Details

The operator takes cytometry data as input and performs the following steps:

1. Loads data from Tercen projections
2. Transforms the data into a format suitable for the ALPODs algorithm
3. Builds a decision tree to classify cell populations
4. Extracts classification rules from the decision tree
5. Returns the classified cell types and their associated rules

The ALPODs algorithm builds a decision tree by recursively splitting the data based on marker expressions. It identifies the optimal splits that maximize information gain at each step. The algorithm stops splitting when one of the following criteria is met:

• Maximum depth is reached
• Minimum number of samples is reached
• Node purity exceeds the threshold
• No further informative splits can be found

References

• ALPODs: Algorithmic Population Descriptions algorithm

See Also

• Tercen
• Tercen R Client