Skip to content

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:

  1. Prepare your input dataset and its corresponding input configuration file.

  2. Set up your computing environment, and create minibatches for subsequent analysis.

  3. Choose the analytical strategy that best suits your project, either LDA or Seurat, to yield clusters or factors from your dataset.

  4. 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

overview_brief Figure 1: A Brief Overview of the Inputs, Outputs, and Process Steps for Pixel-level Analysis.