error with SpatialEnhance

[laetitiacollet]

Dear Edward,

I am trying to run bayes space version ‘1.8.2’ with visium ST data (from Cytassist). When I run
spatialCluster(se_filtered, q=8, platform="Visium", d=7, nit.method="mclust", model="t", gamma=3, nrep=1000, burn.in=100, save.chain=TRUE) it is working.
However with spatialEnhance I have an error message :
se_filtered.enhanced <- spatialEnhance(se_filtered, q=8, d=7, platform="Visium",

•                                 model="t", gamma=3,
  

•                                 jitter_prior=0.3, jitter_scale=3.5,
  

•                                 nrep=1000, #suggest running with at least 10,000 iterations 
  

•                                 burn.in=100,
  

•                                 save.chain=TRUE)
  

Error: subscript contains invalid names

However when I run the same command with platform = "ST" it does work.

Have you already encounter this specific problem ?

Thank you in advance for your reply.

Best regards,

[edward130603]

Sorry for the delayed response. Can you post the output of traceback()? Did you load the data using BayesSpace or another package such as SpatialExperiment?

[laetitiacollet]

Dear Edward, Thank you for your reply. Below you can find my script but I followed the vignette from https://www.ezstatconsulting.com/BayesSpace/articles/BayesSpace.html library(SingleCellExperiment) library(ggplot2) library(BayesSpace) sce <- SingleCellExperiment(assays=list(counts=as(counts, "dgCMatrix")), colData=coord_x_y) ############################### PREPROCESSING DATA ################################ set.seed(102) se_filtered<- spatialPreprocess(sce, platform="Visium", n.PCs=15, n.HVGs=2000) ## allow for log-normalization, findvariable feature and runPCA se_filtered <- qTune(se_filtered, q=seq(2, 15), platform="Visium", d=15, burn.in = 100, nrep = 1000) qPlot(se_filtered) ############################### CLUSTERING DATA WITH BAYES SPACE ################################ set.seed(149) se_filtered <- spatialCluster(se_filtered, q=5, # choose q function of the qplot platform="Visium", d=7,# nb of PC to use init.method="mclust", model="t", gamma=3, #gamma = 2 if platform = ST or 3 if = Visium nrep=1000, #suggest running with at least 10,000 iterations burn.in=100, save.chain=TRUE) # model = 'normal' or 't' but t better performance head(colData(se_filtered)) clusterPlot(se_filtered)

--> This is working well. ############################### ENHANCE RESOLUTION ################################ se_filtered.enhanced <- spatialEnhance(se_filtered, q=8, d=7, platform="Visium", model="t", gamma=3, jitter_prior=0.3, jitter_scale=3.5, nrep=1000, #suggest running with at least 10,000 iterations burn.in=100, save.chain=TRUE) head(colData(se_filtered.enhanced)) clusterPlot(se_filtered.enhanced) --> This does not work and I have the error message : Error: subscript contains invalid names traceback() 12: stop(wmsg(...), call. = FALSE) 11: .subscript_error("subscript contains invalid ", what) 10: NSBS(i, x, exact = exact, strict.upper.bound = !allow.append, allow.NAs = allow.NAs) 9: NSBS(i, x, exact = exact, strict.upper.bound = !allow.append, allow.NAs = allow.NAs) 8: normalizeSingleBracketSubscript(i, xstub) 7: extractCOLS(x, j) 6: extractCOLS(x, j) 5: colData(sce)[position.cols] 4: colData(sce)[position.cols] 3: as.matrix(colData(sce)[position.cols]) 2: .prepare_inputs(sce, use.dimred = use.dimred, d = d, positions = NULL, position.cols = position.cols, xdist = xdist, ydist = ydist) 1: spatialEnhance(se_filtered, q = 8, d = 7, platform = "Visium", model = "t", gamma = 3, jitter_prior = 0.3, jitter_scale = 3.5, nrep = 1000, burn.in = 100, save.chain = TRUE) However, when I replace platform = "ST" instead of Visium, it works. se_filtered.enhanced <- spatialEnhance(se_filtered, q=5, d=7, platform="ST", model="t", gamma=2, jitter_prior=0.3, jitter_scale=3.5, nrep=1000, #suggest running with at least 10,000 iterations burn.in=100, save.chain=TRUE) Calculating labels using iterations 100 through 1000.

traceback()

12: stop(wmsg(...), call. = FALSE) 11: .subscript_error("subscript contains invalid ", what) 10: NSBS(i, x, exact = exact, strict.upper.bound = !allow.append, allow.NAs = allow.NAs) 9: NSBS(i, x, exact = exact, strict.upper.bound = !allow.append, allow.NAs = allow.NAs) 8: normalizeSingleBracketSubscript(i, xstub) 7: extractCOLS(x, j) 6: extractCOLS(x, j) 5: colData(sce)[position.cols] 4: colData(sce)[position.cols] 3: as.matrix(colData(sce)[position.cols]) 2: .prepare_inputs(sce, use.dimred = use.dimred, d = d, positions = NULL, position.cols = position.cols, xdist = xdist, ydist = ydist) 1: spatialEnhance(se_filtered, q = 5, d = 7, platform = "Visium", model = "t", gamma = 2, jitter_prior = 0.3, jitter_scale = 3.5, nrep = 1000, burn.in = 100, save.chain = TRUE)

I remain at your disposal for any additional information you could need. Thanks again for your help, Best regards, Laetitia Collet, MD, PhD student Institut Jules Bordet Medical Oncology Department, Breast Cancer Translational Research Laboratory (BCTL) Université Libre de Bruxelles (ULB) Rue Meylemeersch 90, 1070 Anderlecht Brussels Belgium ***@***.***<https://hubruxelles.be> Lien vers Disclaimer<https://www.hubruxelles.be/sites/default/files/email-disclaimer.html>

…

________________________________ De : Edward Zhao ***@***.***> Envoyé : jeudi 8 août 2024 08:44:55 À : edward130603/BayesSpace Cc : COLLET, Laetitia; Author Objet : Re: [edward130603/BayesSpace] error with SpatialEnhance (Issue #129) Ce message provient de l'extérieur. Soyez attentif lors de l'ouverture de pièces jointes ou de liens provenant d'expéditeurs que vous ne connaissez pas. Deze e-mail komt van een externe afzender. Open geen bijlagen en klik niet op de links, tenzij u de afzender kent. Sorry for the delayed response. Can you post the output of traceback()? Did you load the data using BayesSpace or another package such as SpatialExperiment? — Reply to this email directly, view it on GitHub<#129 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/A4KVCTCF5NALCRRYM674BLTZQMHWPAVCNFSM6AAAAABL33LW6OVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDENZVGA3DSMJYG4>. You are receiving this because you authored the thread.Message ID: ***@***.***>

[edward130603]

Ok, I think this is happening because your colData doesn't have the full spatial set of spatial coordinate info. Your colData should have columns from tissue_positions.csv. Depending on the version of software you are using it would be either 1) the columns row, col, imagerow, and imagecol or 2) the columns array_row, array_col, pxl_row_in_fullres, and pxl_col_in_fullres.

If you used BayesSpace or SpatialExperiment to load in your data, you should have these columns by default. If you used some other tool, you may need to modify your data object to add these columns.

Neighborhood detection for spatialCluster only uses row/col but for spatialEnhance uses both sets of coordinates.