cellengine / cellengine-r-toolkit Goto Github PK

Currently only username/password authentication is supported.

From Federico Gherardini

There's no way to do this from the API currently. A helper function to do this client-side:

getPopByUniqueName <- function(uniqueName) {
  myPops <- getPopulations(experimentId)
  myPops[myPops$uniqueName %in% uniqueName,]
}

deleteGates is an advanced function primarily for deleting tailored gates. Doesn't delete populations. deletePopulationAuto is the normal one to use.

Request from Federico.

Our internal lookupByName function won't work out of the box because it takes an API URL as the first argument; need specializations for each resource type. Some ideas:

lookupByName("gate", "my gate's name")
lookupGateByName("my gate's name")
lookup("gate", list(name = "my gate's name")) # generic lookup fn, probably too verbose

Federico points out that it's not explained clearly in the R docs, and having uncomp as the default is sort of dangerous.

Added in 6bffbbb

Accept everything in the "Optional Body Parameters" of this table as a list.

Right now we have separate lists for X and Y vertices, which is sort of awkward.

+1 @pfgherardini

Missing the compensation (or compensationId) param.

Docs also still say compensatedQ only applies to TSV format.

Deprecated in v0 in #87, to be removed in v1.

We only have getCompensations (plural); need the others:

• getCompensation()
• createCompensation()
• updateCompensation()
• deleteCompensation()

For ease of use, look into splitting these up.

For this HTTP API: https://docs.cellengine.com/api/#download-multiple-files-or-populations-fcs-or-tsv

Probably use readline to prompt for the two-factor code if the first response is a 400 with the message "otp" is required..

in addition to applyScale, we should have a function to apply an entire ScaleSet, like:

applyScaleSet = function(scaleset) {
  cols = lapply(1:nrow(scaleset), function (i) {
    channelName = scaleset[i, "channelName"]
    if (channelName %in% colnames(events)) {
      applyScale(scaleset[i, "scale"], events[, channelName])
    }
  })
  df = as.data.frame(do.call(cbind, cols))
  colnames(df) = colnames(events)
  df
}

also relevant:

col_names = vector(mode="character")
    cols <- vector(mode = "list", length = nrow(scaleset))
    for(i in 1:nrow(scaleset)) {
      channelName = scaleset[i, "channelName"]
      if (channelName %in% colnames(events)) {
        col_names <- append(col_names, channelName)
        cols[[i]] <- applyScale(scaleset[i, "scale"], events[, channelName])
      }
    }
    df = as.data.frame(do.call(cbind, cols))
    colnames(df) = col_names
    out_data = data.matrix(df)

Should default to 0.5.

When you retrieve entities in bulk (i.e. getFcsFiles), and then get a single entity from the list (see code below), then a property that is a list (like annotations) are returned as a nested list.

files = getFcsFiles(id)
f = files[1,]
f$annotations
# prints:
[[1]]
           name value
1 annotations 1     1
2  annotation 2 value

However,

f = getFcsFile(id, byName("myGreatFcsFile")
f$annotations
# prints:
           name value
1 annotations 1     1
2  annotation 2 value

https://github.com/primitybio/cellengine-r-toolkit/blob/b5c93a6f97387887a2917fc92d966511e0c9f7a0/R/getEvents.R#L108

The default behavior is check.names=T, which removes spaces and dashes, making the names awkward.

This is a breaking change. Let's start a v1 branch and gather up these breaking changes in it.

Via @pfgherardini

Since R makes it impossible(?) to create the JSON string needed in some cases.

Alternatively:

> jsonlite::toJSON(list(`$and` = list("A", list(`$not` = "B"))), auto_unbox = T)
{"$and":["A",{"$not":"B"}]} 

Via @pfgherardini

The docs for annotateFcsFile rightly warn that using it will overwrite existing annotations and say to use getFcsFile first. However, getFcsFile[s] returns a list and annotateFcsFile expects a list, so this is tedious to do.

Current method:

fa = data.frame(file$annotations)
annoList = as.list(with(fa, setNames(value, name)))
annoList[["New Anno"]] <- "abcd"

We need to make a cookbook for common tasks. One of the recipes should be for tailoring gates, which is:

res=createPolygonGate(experimentId="5dcf4551b2b26508af8e63b7", "FSC-A", "FSC-H", "my gate", c(100, 200, 300), c(100, 200, 100), tailoredPerFile=TRUE)
res2=createPolygonGate(experimentId="5dcf4551b2b26508af8e63b7", "FSC-A", "FSC-H", "my gate", c(100, 200, 300), c(100, 200, 100), tailoredPerFile=TRUE, fcsFile="Specimen_001_Donor1 24h_001.fcs", gid=res$gate$gid, createPopulation=FALSE)

#51 (comment)

POST /experiment/:experimentId/gates?createPopulation=true does the same thing, better.

getFcsFiles(experimentId, params=list("query"="eq(deleted,3)") returns an empty list, but the same without the query returns a data.frame. This inconsistency makes downstream code complicated. It should always be a data.frame.

byName does not work for deleteGates.R yet.

You could add it inside the if (!is.null(gateId)) block, but it will error out when gates are tailored (which is correct behavior -- tailored gates all have the same name), and when trying to delete compound gates (names vs name). That'd address simple usages, but no pressure to do it.

A better, long-term solution (not this PR) would be a helper function like findGateForFile(fcsFileId=NULL, gid=NULL, name=NULL) helper function that picks a gate by name and FCS file, and supports compound gate names.

It's annoying to have scaleSet$scales return a single-length nested list (see below); it would be better to return as a dataframe. This is a breaking change.

> scaleSet$scales
[[1]] # <- this is a pain
    channelName   scale.type scale.minimum scale.maximum scale.cofactor
1         FSC-A     LogScale             1        100000             NA
2         FSC-W  LinearScale             1        262144             NA
3         SSC-A ArcSinhScale          -200        262144            150
4       Ax488-A ArcSinhScale          -200        262144            150

For plotting FCS files.

It would be nice to have a way to get experiments by name in the functions that require an experimentId parameter, similar to the way you can get parent populations by name when creating gates.

These are currently skipped:

══ Skipped ═════════════════════════════════════════════════════════════════════
1. fromFlowCore accepts `gid` as a splat keyword arg and passes it to the next function (test-convertFromFlowCore.R:6:3) - Reason: flowCore cannot be loaded
...

Interface for this is TBD. It would probably need all of the gates converted first, then passed to toFlowCore for the population conversion.

Right now getGates does not return the fcsFileId column if a gate is tailored but has not actually being tailored yet. This makes working with the return value more complicated. It'd be best to have the fcsFileId column to always be there

The getStatistics method is not accepting the "geometricmean" statistic. All the other ones appear available.

This function should return an individual file according to the description, instead all of them are returned

https://github.com/primitybio/cellengine-r-toolkit/blob/b5c93a6f97387887a2917fc92d966511e0c9f7a0/R/getEvents.R#L58

That's more useful than the default colnames that are V1...VN.

Breaking change, goes in v1.

This would be very helpful for comparing events before and after post-processing.

It would be useful to have getStatistics accept empy arguments for files and populations, defaulting to get statistics for all files and populations

httr::write_disk supports an overwrite option. It'd be good to have getEvents accept an additional overwrite option and pass it down to httr::write_disk