Data Pre-processing

Before SV gene detection, there are several steps to pre-process the data.

Data Filtration

Spots with fewer than ten total counts across all genes are excluded. In addition, genes with more than 90% zero read counts across all spots are dropped.

# filter data
filter.st <- function(count, sample_info, min_total = 10,min_percentage = 0.1){
  gene_num <- ncol(count)
  sample_num <- nrow(count)
  if(sum(rowSums(count) < min_total) == 0){
    if (sum(colSums(count == 0) > (1-min_percentage)*sample_num) > 0){
      sample_f <- sample_info
      count_f <- count[,-(which(colSums(count == 0) > (1-min_percentage)*sample_num))]
    }
    else{
      sample_f <- sample_info
      count_f <- count
    }}
  else{
    if (sum(colSums(count[-which(rowSums(count)<min_total),] == 0) > (1-min_percentage)*sample_num) > 0){
      sample_f <- sample_info[-which(rowSums(count)<min_total),]
      count_f <- count[-which(rowSums(count)<min_total),-(which(colSums(count[-which(rowSums(count)<min_total),] == 0) > (1-min_percentage)*sample_num))]
    }
    else{
      sample_f <- sample_info[-which(rowSums(count)<min_total),]
      count_f <- count[-which(rowSums(count)<min_total),]
    }
  }
  return(list(sample_f,count_f))
}


filter_result <- filter.st(count, sample_info)
count <- filter_result[[2]]
sample_info <- filter_result[[1]]

print(paste0('Dimension of count matrix: ', dim(count)[1],', ', dim(count)[2]))
## [1] "Dimension of count matrix: 260, 11360"

print(paste0('Dimension of location information matrix: ', dim(sample_info)[1],', ', dim(sample_info)[2]))
## [1] "Dimension of location information matrix: 260, 2"

After filtration, the Mouse Olfactory Bulb data (replicate 11) has 260 sample points and 11360 genes.

Size Factor Estimation

Size factor is the input for some SV gene detection approaches. In boost package, we use get.size.factor function to estimate the size factor. The inputs of this function are count matrix \(Y\) and one of the size factor estimation methods (“TSS”, “Q75”, “RLE” and “TMM”). Output is a vector with length \(n\), representing the estimated size factor for spots. In this example, we choose TSS (total sum scaling) as the estimation method.

size_factor <- get.size.factor(count, estimation.method = "TSS")
size_factor <- size_factor/mean(size_factor)

Expression Counts Normalization

Some methods need the normalized gene expression levels as input. normalize.st is the function for obtaining normalized count data, including adjusting for the library size, stablizing the variance, and doing the log-transformation. We provide seven normalization method choices (“TSS”, “Q75”, “RLE”, “TMM”, “A-VST”, “N-VST” and “log-VST”). The output is the normalized expression level matrix, which has the same shape as the input count matrix \(Y\).

normalized_count <- normalize.st(count, scaling.method = "TSS")