Question: Is reference-free spot deconvolution better than reference guided spot deconvolution? What should I use for my analysis?
Answer: There are benefits and limitations to both reference-guided and reference-free methods. Users can evaluate the best methods and tools that fit their needs. Starting in v2.1, Space Ranger automatically performs reference free spot deconvolution for all Visium samples.
- Reference-based or reference guided methods provide the cell type proportions in the tissue based on the single cell reference dataset allowing the user to directly proceed to downstream analysis. They rely on the availability of a high quality single cell dataset that contains cell type annotations. Reference guided spot deconvolution using an annotated single cell reference from a matching tissue sample is the simplest and most accurate method. Unlike reference-free approaches, it does not require prior knowledge of the number of topics (i.e. cell types) or annotating the resulting topics. However, it does require some computational skill to run one of the community developed tools.
- Reference-free methods provide deconvolved proportions of topics which need to be annotated by the user. However, these methods are beneficial when a matched single cell dataset is unavailable. Space Ranger's reference-free spot deconvolution is a great option when a single cell reference is not available or a researcher wants to look at cell type level spatial data using 10x provided software tools. The algorithm is built on latent Dirichlet allocation, similar to STdeconvolve, and outperforms reference guided spot deconvolution that was done using a poorly matched reference. Space Ranger also streamlines the process of picking the number of topics and Loupe makes annotating topics simple by offering an easy to use GUI.
A tutorial for topic annotation using Loupe is available on the support website.
Refer to these articles on the analysis guides page for additional information
- Publication Highlight: Benchmarking Methods to Integrate Spatial and Single-cell Transcriptomics Data
- Integrating 10x Visium and Chromium data with R
- Integrating Single Cell and Visium Spatial Gene Expression Data
- Exploring Your Visium Data: A Spot Deconvolution Story
Products: Spatial Gene Expression
Last updated: Nov 2023