Luis Vazquez, a mathematics major at the University of North Texas, used machine learning algorithms to increase the resolution of kidney slide-seq data.
Spatial transcriptomics is a new technology that spatially maps RNA expression directly from complete tissue slices. Being a part of BSRP 2021 allowed me to work with real biomedical data, learn from great scientists, and meet amazing people. The BSRP coordinators were very supportive, and everyone I met at the Broad was welcoming and fun to work with. I am grateful to have been part of a cohort full of outstanding and inspiring people, and it will be an experience I will never forget.Using Slide-seqV2, a spatial transcriptomics technology, the Greka Lab at the Broad Institute has been able to find differences in cellular organization between healthy and damaged kidney tissue.
However, data from Slide-seqV2 is not at a single-cell resolution, with each data point potentially containing a mix of different cell types. Traditional methods only retain the majority cell type, meaning that some resolution is possibly lost. Since the kidney is structurally complex, we hypothesized that each data spot contains, on average, around 1-3 distinct cell types.
Using the SPOTlight algorithm, we decomposed each data point into its corresponding cell-type proportions. From this decomposed data, we can show how different cell types interact across a kidney tissue sample with higher resolution. In addition, we have also collected data on how individual spots tend to decompose based on the majority cell type. These results suggest that, on average, data spots contain 3-5 distinct cell types.
Furthermore, we have also used K-nearest neighbor filtering to remove noise from the new SPOTlight data. This filtered data shows that important cellular structures are maintained after SPOTlight decomposition. These results corroborate previous findings from the Greka Lab of differences in cellular structures between healthy and damaged kidney tissue in various disease models. In conclusion, SPOTlight can provide a higher resolution perspective on how cell types are organized differently in healthy and damaged kidney tissue.
Project: Using the SPOTlight Algorithm to Increase Resolution of Kidney Slide-Seq Data
Mentors: Jamie Marshall, Kidney Disease Initiative
Anna Greka, Greka Lab