@lcolladotor.bsky.social lcolladotor.github.io lcolladotor.github.io/bioc_team_ds Lessons From Working On The Edge Of Human Brain Spatially-Resolved Transcriptomics Leonardo Collado Torres, LIBD Investigator + Asst. Prof. Johns Hopkins Biostatistics Spatial Biology West Coast US 2024 December 5th 2024 Slides available at speakerdeck.com/lcolladotor

10.1016/j.biopsych.2020.06.005 Michael Gandal @mikegandal Transcriptomic Insight Into the Polygenic Mechanisms Underlying Psychiatric Disorders

Background: Human DLPFC 3 Louise A Huuki-Myers @lahuuki.bsky.social

Zoom in: spatial omics Kristen R Maynard @kr_maynard Keri Martinowich @martinowk Stephanie C Hicks @stephaniehicks Andrew E Jaffe @aejaffe Stephanie C Page @cerceopage

Visium Platform for Spatial Gene Expression Image from 10x Genomics - A slide contains 4 capture areas, each full of thousands of 55um-wide “spots” (often containing 1-10 cells) - Unique barcodes in each spot bind to particular genes; after sequencing, gene expression can be tied back to exact spots, forming a spatial map Kristen R. Maynard 5

#HumanPilot study explainer video youtu.be/HGioWKuI3ek https://lcolladotor.github.io/2024/05/23/humanpilot-first- spatially-resolved-transcriptomics-study-using-visium/

2 pairs spatial adjacent replicates x subject = 12 sections 7 Subject 1 Subject 2 Subject 3 Adjacent spatial replicates (0μm) Adjacent spatial replicates (300μm) PCP4 10.1038/s41593-020-00787-0

“Pseudo-bulking” collapses data: spot to layer level 8 10.1038/s41593-020-00787-0

DOI: 10.1038/s41593-020-00787-0 twitter.com/lcolladotor/status/1233661576433061888 from 2020-02-29 twitter.com/CrowellHL/status/ 1597579271945715717 DOI 10.1093/bioinformatics/btac299 Since Feb 2020 spatialLIBD::fetch_data() provides access to SpatialExperiment R/Bioconductor objects

Some challenges ⚠ with Visium 10

Some analysis differences ● 10x Genomics CellRanger → SpaceRanger (or other tools) ● DropletUtils::read10xVisium() → spatialLIBD::read10xVisiumWrapper() ● Spot QC ○ Drop high mitochondria % spots? ○ Drop spots with low UMI counts? ■ Biological vs technical reasons ● Spot clustering (spatially-aware) and/or manual annotation ● Choosing a spot clustering resolution ○ Can be guided by spatially cluster registration

Some analysis differences ● 10x Genomics CellRanger → SpaceRanger (or other tools)

Align your data with spaceranger --count 13 www.10xgenomics.com/support/software/space-ranger/latest

Loupe Browser Manually: ● Align fiducial frame ● Select spots “in tissue” lmweber.org/Visium- data-preprocessing/

bioconductor.org/packages/spatialLIBD Pardo et al, 2022 DOI 10.1186/s12864-022-08601-w Maynard, Collado-Torres, 2021 DOI 10.1038/s41593-020-00787-0 Brenda Pardo Abby Spangler @PardoBree @abspangler Louise A. Huuki-Myers @lahuuki

spatialLIBD::vis_*() functions

Some analysis differences ● 10x Genomics CellRanger → SpaceRanger (or other tools) ● DropletUtils::read10xVisium() → spatialLIBD::read10xVisiumWrapper()

spatialLIBD::read10xVisiumWrapper() 18 bioconductor.org/packages/spatialLIBD

spatialLIBD::run_app() 19 libd.shinyapps.io/Habenula_Visium_raw/ Default non-spatially aware graph based clustering results + K-means (k = 2 to 10)

Some analysis differences ● 10x Genomics CellRanger → SpaceRanger (or other tools) ● DropletUtils::read10xVisium() → spatialLIBD::read10xVisiumWrapper() ● Spot QC ○ Drop high mitochondria % spots? ○ Drop spots with low UMI counts?

sc/snRNA-seq QC metrics such as # detected genes, # UMI, mitochondria expression % are likely biologically related! https://lcolladotor.github.io/2024/05/23/humanpilot-first- spatially-resolved-transcriptomics-study-using-visium/

Some analysis differences ● 10x Genomics CellRanger → SpaceRanger (or other tools) ● DropletUtils::read10xVisium() → spatialLIBD::read10xVisiumWrapper() ● Spot QC ○ Drop high mitochondria % spots? ○ Drop spots with low UMI counts? ■ Biological vs technical reasons

scuttle::isOutlier(sce.416b$sum, log=TRUE, type="lower") drive.google.com/drive/folders/1cn5d-Ey7-kkMiex8-74qxvxtCQT6o72h Slide by Peter Hickey

10.1089/genbio.2023.0019 Fig S9 Low library size (UMIs, # detected genes): ● Biological: ○ Not all cells / layers are equally active ● Technical: ○ Edge of tissue: should have been called “out of tissue” ○ Truly low quality

● Spot-level QC ○ Define a local neighborhood using k-nearest neighbors based on spatial coordinates ○ Use neighborhood to calculate z-score for the QC metric (library size, detected genes, mito rate) ○ Identity outlier spots based on z-score Fig 1: An overview of SpotSweeper SpotSweeper by Totty, Hicks, Guo 10.1101/2024.06.06.597765

Fig 3: SpotSweeper detects consistent set of spots with systematically low library size driven by barcode biases. ● 6 barcodes in Visium slides have low library sizes across datasets ● Synthetic barcodes at these spots may be responsible for downstream quality issues ○ Certain k-mers show bias towards larger or smaller library sizes SpotSweeper by Totty, Hicks, Guo 10.1101/2024.06.06.597765

Prashanthi Ravichandran @prashanthi-ravichandran (GH) Artifacts in general are normalized away by library size, though there are caveats 10.1126/science.adh1938

Diffusion issues: could be related to permeabilization step 10.1126/science.adh1938

Having more data is useful to provide context! Here 4 new samples have low sequencing saturation (outliers) but are within range of good samples from other studies

Having more data is useful to provide context! Those 4 samples have great median UMI counts per spot ^_^

Some analysis differences ● 10x Genomics CellRanger → SpaceRanger (or other tools) ● DropletUtils::read10xVisium() → spatialLIBD::read10xVisiumWrapper() ● Spot QC ○ Drop high mitochondria % spots? ○ Drop spots with low UMI counts? ■ Biological vs technical reasons ● Spot clustering (spatially-aware) and/or manual annotation

Visualize known marker genes

Visualize known marker genes: manually annotate

34 bioconductor.org/packages/spatialLIBD Visualizing multiple genes

35 bioconductor.org/packages/spatialLIBD Spatially-aware clustering results BayesSpace PRECAST …

Some analysis differences ● 10x Genomics CellRanger → SpaceRanger (or other tools) ● DropletUtils::read10xVisium() → spatialLIBD::read10xVisiumWrapper() ● Spot QC ○ Drop high mitochondria % spots? ○ Drop spots with low UMI counts? ■ Biological vs technical reasons ● Spot clustering (spatially-aware) and/or manual annotation ● Choosing a spot clustering resolution ○ Can be guided by spatially cluster registration

37 bioconductor.org/packages/spatialLIBD 10.1126/science.adh1938

spatialLIBD::registration_wrapper() + layer_stat_cor() 38 10.1126/science.adh1938

spatialLIBD interactive apps ● Useful for visualizing & annotating the data ● Immediately make them after running SpaceRanger ● Update them as we generate more results ● Share them with the publication of the results ● spatial.libd.org/spatialLIBD/ ● research.libd.org/spatialDLPFC/#interactive-websites ● research.libd.org/Visium_SPG_AD/#interactive-websites ● …

Getting our feet wet with Visium HD 40 @Nick-Eagles Nicholas J. Eagles Kristen R Maynard @kr_maynard Kelsey D. Montgomery Sarah E. Maguire

Visium HD Overview - Visium HD is composed of square bins, not spots, at 3 different resolutions - 8μm bin size is recommended for analysis by 10x Genomics - Compare to 100μm distance between spots! - ~700k bins, compared to ~5k spots! - 2μm bin size is subcellular, and can be combined with segmentation to form cells “Bin level”: 8μm bins “Cell level”: Individual cells from combining 2μm bins 16μm bin border 8μm bin border 2μm bin border cell

Top SVGs (bin level) 1. Lower resolution with SEraster 10.1093/bioinformatics/btae412 2. Compute SVGs with nnSVG 10.1038/s41467-023-39748-z 3. Visualize with spatialLIBD 10.1186/s12864-022-08601-w 1 5 9

1 5 9 1. Compute HVGs with scran 10.12688/f1000research.9501.2 2. Visualize with spatialLIBD 10.1186/s12864-022-08601-w Top HVGs (cell level)

Spatially informed cell clustering with MERINGUE 10.1101/gr.271288.120

Banksy: cell-level spatial domains (k = 2, 4, 8) 10.1038/s41588-024-01664-3

FICTURE clustering * Note: mirrored and rotated data 10.1101/2023.11.04.565621

Software keeps evolving and as leaders in the field we aim to use the best methods 47 Moses, L., Pachter, L. Museum of spatial transcriptomics. Nat Methods 19, 534–546 (2022). https://doi.org/10.1038/s41592-022-01409-2

Visium HD Software Comparison Software tool Runs on our data Run time (per sample) Required memory (per sample, GB) Integrates well with SpatialExperiment Operates on multiple samples simultaneously Uses GPU Documentation quality Authors responsive bin2cell Yes 30 min < 32 Yes No No Good Yes FICTURE Yes 3 hours < 64 No No No Okay Yes SEraster Yes 20 min < 64 Yes No No Good Yes MERINGUE Sort of 1 day > 400 Yes No No Okay Yes Giotto Yes Varied Varied Yes Yes No Needs improvement Banksy Yes 3 hours > 250 Yes Yes No Okay HERGAST Yes Okay ENACT @Nick-Eagles Nicholas J. Eagles #RStats and #PyStats https://github.com/LieberInstitute/jhpce_module_source

Learn in public Learn continuously Share what you’ve learned youtube.com/@lcolladotor

50 research.libd.org/rstatsclub LIBD rstats club videos

51 research.libd.org/rstatsclub LIBD rstats club videos

@MadhaviTippani Madhavi Tippani @HeenaDivecha Heena R Divecha @lmwebr Lukas M Weber @stephaniehicks Stephanie C Hicks @abspangler Abby Spangler @martinowk Keri Martinowich @CerceoPage Stephanie C Page @kr_maynard Kristen R Maynard @lcolladotor Leonardo Collado-Torres @Nick-Eagles Nicholas J Eagles Kelsey D Montgomery Sang Ho Kwon Image Analysis Expression Analysis Data Generation Thomas M Hyde @lahuuki Louise A Huuki-Myers @BoyiGuo Boyi Guo @mattntran Matthew N Tran @sowmyapartybun Sowmya Parthiban Slides available at speakerdeck.com /lcolladotor + Many more LIBD, JHU, and external collaborators @mgrantpeters Melissa Grant-Peters @prashanthi-ravichandran (GH) Prashanthi Ravichandran

lcolladotor.github.io @lcolladotor.bsky.social Slides: