fictureR

fictureR: FICTURE from Si et al. 2024 in R

fictureR is an R implementation of the FICTURE method from Si et al. 2024, designed to efficiently handle high-resolution spatial transcriptomics (ST) data. This package focuses on variational Expectation-Maximization (EM) steps and attempts to perform Latent Dirichlet Allocation (LDA) training for analyzing complex, segmentation-free ST datasets at submicron resolution.

Purpose

Spatial transcriptomics (ST) data allow for transcriptome-wide gene expression analysis with high spatial precision, often at submicron resolution. Traditional methods rely on cell segmentation or gridding, which often fail in tissues with irregular and diverse cell sizes and shapes. FICTURE bypasses these issues with a segmentation-free spatial factorization approach that efficiently scales to large datasets, including those with billions of spatial coordinates.

The R implementation focuses on: - Performing the variational EM step, crucial for spatial factorization. - Attempting LDA training to model the spatial distribution of gene expression in tissues.

The package is ideal for users working with both sequencing- and imaging-based ST data and is compatible with large-scale, high-resolution studies

Installation

You can install the development version of fictureR from Github with:

# install.packages("devtools")
devtools::install_github("walterwilliamson/fictureR")

Example

Here we show the basic usage of the slda_decoder and visualizer functions, using an example dataset of mouse liver gene expression from [Vizgen MERFISH]:(https://info.vizgen.com/mouse-liver-access)

Due to the large size of the dataset and slow nature of our function, we have used a subset of the data for demonstration purposes.

library(fictureR)
#> Warning: replacing previous import 'dplyr::count' by 'matrixStats::count' when
#> loading 'fictureR'
library(dplyr)
#> 
#> Attaching package: 'dplyr'
#> The following objects are masked from 'package:stats':
#> 
#>     filter, lag
#> The following objects are masked from 'package:base':
#> 
#>     intersect, setdiff, setequal, union
mouse_anchors <- fictureR::mouse_anchors
mouse_transcripts <- fictureR::mouse_transcripts

# Visualization of anchors before variational EM, subsetting to a small region
visualizer(mouse_anchors %>% filter(x >= 6690.1, x <= 6790.1,
                                    y >= 6772, y <= 6872), 
           type = c("anchor"))

# Decoding input data
# mouse_output <- slda_decoder(mouse_transcripts, mouse_anchors, scale = 100)

# Updated anchor assignments following decoding
visualizer(fictureR::mouse_updated_anchor, type = c("anchor"))

# 
# Visualization of pixel data following decoding
visualizer(fictureR::mouse_pixel, type = c("pixel"))