Usage

Installation

To use TriTan, first install it using pip:

(.venv) $ pip install TriTan

Import TriTan as:

import TriTan

Initialise a TriTan model:

tritan = TriTan.TriTan()

The input to TriTan is a MuData object. Train the model:

tritan.fit(mdata)

Parameter	Description
n_modality	int, (default=2). The number of the modalities.
resolution	float, (default=0.6). The resolution parameter, ranging from 0 to 1, controls the granularity of the clusters. Specifically, it adjusts the size of the clusters: Higher resolution values lead to more and smaller clusters, encouraging the algorithm to find finer structures in the data. Lower resolution values result in fewer and larger clusters, merging more points into broader groups.
precomputed	bool, (default=False). TriTan uses SVD to enhance the speed of factorization. When precomputed is set to False, TriTan will run the SVD factorization within the model. When precomputed is set to True, users can provide their own SVD factorization to save the time.
svd_mods	When precomputed is set to True, uses should use svd_mods to input their own svd factorization as the format of dictionary: >>> u_rna, s_rna, v_rna = randomized_svd(X_gene, n_components=300, random_state=0) >>> u_atac, s_atac, v_atac = randomized_svd(X_atac, n_components=300, random_state=0) >>> svd_mods = {'rna': [u_rna, v_rna], 'atac': [u_atac, v_atac]}
n_component	dic, (default={‘rna’: [20,50], ‘atac’: [20,50]}). Specifies the number of components for SVD decomposition in the cell and feature spaces for each modality. The number of components can be adjusted based on the dataset and corresponding singular value scree plots. The parameter allows for user-defined SVD decomposition for each modality.