Pixel-level Analysis¶
This section provides an example of how to identify spatial factors at pixel-level resolution in spatial transcriptomics data.
Analytical Strategies¶
NEDA currently offers two analytical strategies:
1) Latent Dirichlet Allocation (LDA) + FICTURE: This strategy utilizes Latent Dirichlet Allocation (LDA) to identify spatial factors, and then uses FICTURE to map these factors onto a histological space with pixel-level resolution.
2) Seurat + FICTURE: This strategy uses multi-dimensional clustering via Seurat to explore cell type clusters and then projects those clusters into a histological space using FICTURE, achieving pixel-level resolution.
A Step-by-Step Procedure¶
Before beginning the analysis, ensure that NEDA and its dependencies are installed properly. Then, follow these steps as outlined:
-
Prepare your input dataset and its corresponding input configuration file.
-
Set up your computing environment, and create minibatches for subsequent analysis.
-
Choose the analytical strategy that best suits your project, either LDA or Seurat, to yield clusters or factors from your dataset.
-
Transform and decode these clusters or factors on your input data at pixel-level resolution.
Each step contains detailed instructions for:
- the purpose of each step;
- the execution commands;
- necessary input and output files;
- definitions of auxiliary parameters, as outlined in the scripts for each step.
An Overview¶
Figure 1: A Brief Overview of the Inputs, Outputs, and Process Steps for Pixel-level Analysis.