Hi, 

I am using Scran and the tutorial as below to analyze singe-cell ATAC-seq data from 10x. https://www.bioconductor.org/packages/devel/workflows/vignettes/simpleSingleCell/inst/doc/work-3-tenx.html#7_modelling_the_mean-variance_trend

When using the example data, I could run it without any issue.

But when using my own data, at the step of fitting I got the following error:

>new.trend <- makeTechTrend(x=sce)
Error in seq.default(from = 0, to = upper.value, length.out = 100) : 
  'to' must be a finite number

Could anyone kindly help me on this? Thank you very much! My sessionInfo() is:

R version 3.5.1 (2018-07-02)
Platform: x86_64-apple-darwin15.6.0 (64-bit)
Running under: macOS High Sierra 10.13.6

Matrix products: default
BLAS: /System/Library/Frameworks/Accelerate.framework/Versions/A/Frameworks/vecLib.framework/Versions/A/libBLAS.dylib
LAPACK: /Library/Frameworks/R.framework/Versions/3.5/Resources/lib/libRlapack.dylib

locale:
[1] en_US.UTF-8/en_US.UTF-8/en_US.UTF-8/C/en_US.UTF-8/en_US.UTF-8

attached base packages:
[1] parallel  stats4    stats     graphics  grDevices utils     datasets  methods   base     

other attached packages:
 [1] scran_1.9.16                EnsDb.Mmusculus.v79_2.99.0  ensembldb_2.5.4            
 [4] AnnotationFilter_1.5.2      GenomicFeatures_1.33.2      AnnotationDbi_1.43.1       
 [7] scater_1.9.14               ggplot2_3.0.0               DropletUtils_1.1.8         
[10] SingleCellExperiment_1.3.10 SummarizedExperiment_1.11.6 DelayedArray_0.7.28        
[13] matrixStats_0.54.0          Biobase_2.41.2              GenomicRanges_1.33.13      
[16] GenomeInfoDb_1.17.1         IRanges_2.15.16             S4Vectors_0.19.19          
[19] BiocGenerics_0.27.1         BiocParallel_1.15.8        

The upper.value refers to the maximum of the per-row average of the log-normalized values in logcounts(sce). The error is telling you that upper.value is not finite, which suggests that you have NA values in logcounts(sce). How these came about is anyone's guess, though the most likely explanation is that there are non-positive size factors somewhere in your analysis.

I haven't dealt with scATAC-seq data so I wouldn't know its peculiarities. But, I have worked with both ChIP-seq and scRNA-seq, and both of them require some care when performing scaling normalization. I would imagine that scATAC-seq "inherits" both sets of difficulties from the two technologies.