Align features refactoring (vol. 2) #88
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -1,20 +1,78 @@ | ||
| to_attach <- function(pick, number_of_samples, use = "sum") { | ||
| strengths <- rep(0, number_of_samples) | ||
| if (is.null(nrow(pick))) { | ||
| # this is very strange if it can ever happen | ||
| # maybe commas are missing? we want the same as below | ||
| # but also why if there are no rows... | ||
| strengths[pick[6]] <- pick[5] | ||
| return(c(pick[1], pick[2], pick[1], pick[1], strengths)) | ||
| } else { | ||
| for (i in seq_along(strengths)) { | ||
| # select all areas from the same sample | ||
| areas <- pick[pick[, 6] == i, 5] | ||
| if (use == "sum") | ||
| strengths[i] <- sum(areas) | ||
| if (use == "median") | ||
| strengths[i] <- median(areas) | ||
| # can be NA if pick does not contain any data from a sample | ||
| } | ||
| # average of m/z, average of rt, min of rt, max of rt, sum/median of areas | ||
| return(c(mean(pick[, 1]), mean(pick[, 2]), min(pick[, 1]), | ||
| max(pick[, 1]), strengths)) | ||
| } | ||
| } | ||
|
|
||
|
|
||
| select_rt <- function(sample, rt_tol_relative, min_occurrence, number_of_samples) { | ||
| # Kernel Density Estimation, this time for retention time | ||
| den <- density(sample[, 2], bw = rt_tol_relative / 1.414) | ||
| # again select statistically significant points | ||
| turns <- find.turn.point(den$y) | ||
| peaks <- den$x[turns$pks] | ||
| valleys <- den$x[turns$vlys] | ||
| for (i in seq_along(peaks)) { | ||
| lower_bound <- max(valleys[valleys < peaks[i]]) | ||
| upper_bound <- min(valleys[valleys > peaks[i]]) | ||
| # select data with m/z within lower and upper bound from density estimation | ||
| selected <- which(sample[, 2] > lower_bound & sample[, 2] <= upper_bound) | ||
| sample_grouped <- sample[selected, ] | ||
| if (!is.null(nrow(sample_grouped))) { | ||
| if (length(unique(sample_grouped[, 6])) >= min_occurrence) { | ||
| # continue if data is still from at least 'min_occurrence' samples | ||
| return(c(to_attach(sample_grouped, number_of_samples, use = "sum"), | ||
| to_attach(sample_grouped[, c(1, 2, 3, 4, 2, 6)], number_of_samples, use = "median"), | ||
| sd(sample_grouped[, 1], na.rm = TRUE) | ||
| ) | ||
| ) | ||
| } | ||
| } | ||
| } | ||
| } | ||
|
|
||
|
|
||
| select_mz <- function(sample, mz_tol_relative, rt_tol_relative, min_occurrence, number_of_samples) { | ||
| # Kernel Density Estimation with target standard deviation 'mz_tol_relative' multiplied by median of m/z | ||
| des <- density(sample[, 1], bw = mz_tol_relative * median(sample[, 1])) | ||
| # find the peaks and valleys of a density function | ||
| turns <- find.turn.point(des$y) | ||
| peaks <- des$x[turns$pks] | ||
| valleys <- des$x[turns$vlys] | ||
| for (i in seq_along(peaks)) { | ||
| lower_bound <- max(valleys[valleys < peaks[i]]) | ||
| upper_bound <- min(valleys[valleys > peaks[i]]) | ||
| # select data with m/z within lower and upper bound from density estimation | ||
| selected <- which(sample[, 1] > lower_bound & sample[, 1] <= upper_bound) | ||
| sample_grouped <- sample[selected, ] | ||
| if (!is.null(nrow(sample_grouped))) { | ||
| # continue if data is still from at least 'min_occurrence' samples | ||
| if (length(unique(sample_grouped[, 6])) >= min_occurrence) { | ||
| return(select_rt(sample_grouped, rt_tol_relative, min_occurrence, number_of_samples)) | ||
| } | ||
| } | ||
| } | ||
| } | ||
|
|
||
|
|
||
| # returns a list of aligned features and original peak times | ||
| feature.align <- function(features, | ||
| min_occurrence = 2, | ||
|
|
@@ -34,22 +92,22 @@ feature.align <- function(features, | |
| if (number_of_samples > 1) { | ||
| values <- get_feature_values(features, rt_colname) | ||
| mz_values <- values$mz | ||
| rt <- values$rt | ||
| sample_id <- values$sample_id | ||
|
|
||
| # sort all values by m/z, if equal by rt | ||
| ordering <- order(mz_values, rt) | ||
| mz_values <- mz_values[ordering] | ||
| rt <- rt[ordering] | ||
| sample_id <- sample_id[ordering] | ||
|
Comment on lines
+135
to
+139
There was a problem hiding this comment. Data could be arranged in a |
||
|
|
||
| # find relative m/z tolerance level | ||
| if (is.na(mz_tol_relative)) { | ||
| mz_tol_relative <- find.tol(mz_values, mz_max_diff = mz_max_diff, do.plot = do.plot) | ||
| if (length(mz_tol_relative) == 0) { | ||
| mz_tol_relative <- 1e-5 | ||
| warning("Automatic tolerance finding failed, 10 ppm was assigned. | ||
| May need to manually assign alignment mz tolerance level.") | ||
|
Comment on lines
+146
to
+147
There was a problem hiding this comment. This is something that we should address maybe: I doubt someone will see this warning given that most of our test-cases produce dozens of warnings. There was a problem hiding this comment. Do you think the function should fail with an error? Or we could parametrise the whole function to use default value vs. fail with an error. There was a problem hiding this comment. No, I don't think an error would be an appropriate solution. I just wanted to point out that apLCMS may produce a meaningful warning, so we should make sure that the warning doesn't get lost in a bunch of deprecation warnings, etc. I think we just have to reduce the number of other warnings like deprecation, imports, and so on. Should be done with #93. |
||
| } | ||
| } else if (do.plot) { | ||
| draw_plot(main = "alignment m/z tolerance level given", | ||
|
|
@@ -62,119 +120,107 @@ feature.align <- function(features, | |
| } | ||
|
|
||
| # find relative retention time tolerance level | ||
| # also does some preprocessing grouping steps | ||
| all_features <- find.tol.time(mz_values, | ||
| rt, | ||
| sample_id, | ||
| number_of_samples = number_of_samples, | ||
| mz_tol_relative = mz_tol_relative, | ||
| rt_tol_relative = rt_tol_relative, | ||
| mz_tol_absolute = mz_tol_absolute, | ||
| do.plot = do.plot) | ||
| rt_tol_relative <- all_features$rt.tol | ||
|
|
||
| message("**** performing feature alignment ****") | ||
| message(paste("m/z tolerance level: ", mz_tol_relative)) | ||
| message(paste("time tolerance level:", rt_tol_relative)) | ||
|
|
||
| # create zero vectors of length number_of_samples + 4 ? | ||
| aligned_features <- pk.times <- rep(0, 4 + number_of_samples) | ||
| mz_sd <- 0 | ||
|
|
||
| labels <- unique(all_features$grps) | ||
| area <- grps <- mz_values | ||
|
|
||
| # grouping the features based on their m/z values (assuming the tolerance level) | ||
| sizes <- c(0, cumsum(sapply(features, nrow))) | ||
| for (i in 1:number_of_samples) { | ||
| sample <- features[[i]] | ||
| # order by m/z then by rt | ||
| sample <- sample[order(sample[, 1], sample[, 2]),] | ||
|
|
||
| # select preprocessed features belonging to current sample | ||
| group_ids <- which(all_features$lab == i) | ||
| # select m/z, rt and their group ID | ||
| sample_grouped <- cbind(all_features$mz[group_ids], all_features$rt[group_ids], all_features$grps[group_ids]) | ||
| # order them again? should be ordered already... | ||
| sample_grouped <- sample_grouped[order(sample_grouped[, 1], sample_grouped[, 2]),] | ||
|
There was a problem hiding this comment. Did you try to run tests without this line? There was a problem hiding this comment. Good point, without that the test case fails, so I guess they weren't really ordered... |
||
|
|
||
| # update m/z, rt, area values with ordered ones | ||
| mz_values[(sizes[i] + 1):sizes[i + 1]] <- sample[, 1] | ||
| rt[(sizes[i] + 1):sizes[i + 1]] <- sample[, 2] | ||
| area[(sizes[i] + 1):sizes[i + 1]] <- sample[, 5] | ||
| # assign row identifier | ||
| grps[(sizes[i] + 1):sizes[i + 1]] <- sample_grouped[, 3] | ||
| # assign batch identifier | ||
| sample_id[(sizes[i] + 1):sizes[i + 1]] <- i | ||
| } | ||
|
|
||
| # table with number of values in a group | ||
| groups_cardinality <- table(all_features$grps) | ||
| # count those with minimal occurrence | ||
| # (times 3 ? shouldn't be number of sample) !!! | ||
| curr.row <- sum(groups_cardinality >= min_occurrence) * 3 | ||
| mz_sd <- rep(0, curr.row) | ||
|
|
||
| sel.labels <- as.numeric(names(groups_cardinality)[groups_cardinality >= min_occurrence]) | ||
|
|
||
| # retention time alignment | ||
| aligned_features <- | ||
| foreach(i = seq_along(sel.labels), .combine = rbind) %do% { | ||
| if (i %% 100 == 0) | ||
| gc() # call Garbage Collection for performance improvement? | ||
| # select a group | ||
| group_ids <- which(grps == sel.labels[i]) | ||
| if (length(group_ids) > 1) { | ||
| # select data from the group | ||
| sample <- cbind(mz_values[group_ids], rt[group_ids], rt[group_ids], rt[group_ids], area[group_ids], sample_id[group_ids]) | ||
| # continue if data is from at least 'min_occurrence' samples | ||
| if (length(unique(sample[, 6])) >= min_occurrence) { | ||
| return(select_mz(sample, mz_tol_relative, rt_tol_relative, min_occurrence, number_of_samples)) | ||
| } | ||
| } else { | ||
| if (min_occurrence == 1) { | ||
| sample_grouped <- c(mz_values[group_ids], rt[group_ids], rt[group_ids], rt[group_ids], area[group_ids], sample_id[group_ids]) | ||
| return(c(to_attach(sample_grouped, number_of_samples, use = "sum"), | ||
| to_attach(sample_grouped[c(1, 2, 3, 4, 2, 6)], number_of_samples, use = "median"), | ||
| NA | ||
| ) | ||
| ) | ||
| } | ||
| } | ||
| return(NULL) | ||
| } | ||
|
|
||
| # select columns: average of m/z, average of rt, min of m/z, max of m/z, median of rt per sample (the second to_attach call) | ||
| pk.times <- aligned_features[, (5 + number_of_samples):(2 * (4 + number_of_samples))] | ||
| mz_sd <- aligned_features[, ncol(aligned_features)] | ||
| # select columns: average of m/z, average of rt, min of m/z, max of m/z, sum of areas per sample (the first to_attach call) | ||
| aligned_features <- aligned_features[, 1:(4 + number_of_samples)] | ||
|
Comment on lines
+238
to
+241
There was a problem hiding this comment. Computations for the numbers of columns could be extracted to variables to clearly indicate what is being used and why. There was a problem hiding this comment. This will be changed in #87 anyway. |
||
|
|
||
| # rename columns on both tables, samples are called "exp_i" | ||
| colnames(aligned_features) <- | ||
| colnames(pk.times) <- c("mz", "time", "mz.min", "mz.max", paste("exp", 1:number_of_samples)) | ||
|
|
||
| # return both tables and both computed tolerances | ||
| rec <- new("list") | ||
| rec$aligned.ftrs <- aligned_features | ||
| rec$pk.times <- pk.times | ||
| rec$mz.tol <- mz_tol_relative | ||
| rec$chr.tol <- rt_tol_relative | ||
|
|
||
| if (do.plot) { | ||
| hist(mz_sd, xlab = "m/z SD", ylab = "Frequency", | ||
| main = "m/z SD distribution") | ||
| hist(apply(pk.times[, -1:-4], 1, sd, na.rm = TRUE), | ||
| xlab = "Retention time SD", ylab = "Frequency", | ||
|
Comment on lines
254
to
258
There was a problem hiding this comment. This could be moved to the plotting. There was a problem hiding this comment. Same as in @zargham-ahmad's comment, do you have a suggestion? I think we would just create an artificial function which takes the same arguments as
Comment on lines
+255
to
258
There was a problem hiding this comment. Do you have any suggestions? I think we would just create an artificial function which takes the same arguments as |
||
|
|
||
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I guess the problem lies there that nrow(pick) might be
nullif pick has the wrong type - meaning it is not a table but a list or so and then it actually behaves differently.There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Exactly. That also happens with
nrowcheck inprof.to.features.There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Okay, that makes sense, it was just suspicious for me.