gszep/flowatlas.jl

an interactive explorer for flow cytometry data

flowatlas.jl's Issues

Add Gating-ML files for the rest of FCS files with B-cells

Time series of all channels signal anomaly gating

@mad3cientist upload 2 examples with images showing manual anomaly gating.

filters do not show cells when no conditions/groups are present

missing group="tLN" for sampleIDs 9 and 34

The metadata from the workspace file currently shows

       <Group name="tLN" >
         <Criteria>
           <Criterion keyword="TUBE NAME"  function="Contains"  value="tln" />
         </Criteria>
         <ManuallyIncludedSamples>
           <SampleRef sampleID="11" />
         </ManuallyIncludedSamples>
         <SampleRefs>
           <SampleRef sampleID="33" />
           <SampleRef sampleID="25" />
           <SampleRef sampleID="11" />
         </SampleRefs>
         <Keywords/>
       </Group>

Singlet and Live cell extraction

Density thresholding on FSC-W vs FSC-A then on SSC-W vs SSC-A to get singlets.
Then on Z-UV negative to get live cells. @mad3cientist Upload 2 examples into a separate folder under data/testing

update to use JSServe v2

I am several versions behind JSServe and would probably benefit from all the new functionality

Open a set of FCS files without a .wsp file?

Cool project! I was wondering if it was possible to open a folder with .fcs files without a .wsp file?

Something like:

julia> readdir()
5-element Vector{String}:
 "Specimen_001_1 BFP_002.fcs"
 "Specimen_001_1 CTL_001.fcs"
 "Specimen_001_2 BFP_004.fcs"
 "Specimen_001_2 CTL_003.fcs"

julia> FlowAtlas.run(""; files = "*.fcs")
[ Info: Loading FCS files...
ERROR: AssertionError: no such file: 
Stacktrace:
 [1] load(path::String; files::String, cols::Symbol, kwargs::@Kwargs{})
   @ FlowWorkspace ~/.julia/packages/FlowWorkspace/8HLjp/src/Utils.jl:32
 [2] load
   @ ~/.julia/packages/FlowWorkspace/8HLjp/src/Utils.jl:30 [inlined]
 [3] run(path::String; files::String, port::Int64, url::String, cols::Symbol, drop::Vector{…}, nlevels::Int64, nbins::Int64, heatmapQuantileEdge::Float64, violinQuantileEdge::Float64, channelScheme::ColorSchemes.ColorScheme{…}, labelScheme::ColorSchemes.ColorScheme{…}, hold::Bool)
   @ FlowAtlas ~/.julia/packages/FlowAtlas/yxybS/src/FlowAtlas.jl:65
 [4] top-level scope
   @ REPL[22]:1
Some type information was truncated. Use `show(err)` to see complete types.

My package environment on Julia 1.10:

(jl_TZ3iwY) pkg> st
Status `/tmp/jl_TZ3iwY/Project.toml`
  [ef7debb3] FlowAtlas v0.1.13

I would love to use this tool to visualize my populations and draw gates, but I understand if that's out of the scope of this project.

update violin plots when dragging gates

csv, svg export of panel figures

Optimising cluster hyperparameters

For example, TRegs can be split into +/- CD69. @mad3cientist upload into comments examples of subpopulation that have not been split properly. Report metadata on number of merged clusters that have been assigned to one label and have the option of exploring unmatched clusters. Relation of ROC curve? Perhaps we would like to explore ROC vs cluster number.

batch effect warning

if a measure of mixing between groups for labelled populations goes below a certain threshold, throw a warning to the user notifying them about which samples and which population contains the batch effect.

Explore other clustering algorithm options

Initial list here:
https://scikit-learn.org/stable/modules/clustering.html

If still unsatisfactory then do:

user input for naming polygons; display in violin plot legend

restructure runner as julia package

Benchmark against FlowCAP-III

Confusion matrix and ROC

Parse GateML to generate labels, compare labels to predictions
Visualise accuracies using confusion matrix, generate precision recall curve (ROC-curve) by changing confidence thresholds in profile projection

labels on colour bar showing fluorescence heatmap are clipped

I will move the colour bar a couple of pixels to the left

use quadtree to store embedding coordinates

For more efficient data access when rendering tiles

warning when user compares results from differing data origins

We would like warnings for the following two cases

• two different lasers that give the same marker name [hard warning]
• two different gating strategies that give the same population label [soft warning]

colorby populations, conditions, groups

embedding with missing data

we need to decide what to do about missing data. Substituting zeros in the channel columns seems arbitrary.. what is the appropriate way of dealing with this?

use browser cache to store default colour scheme preferences

error right at the beginning

Hello,
I'm getting this error:

What should I do?

Many thanks,
Vincent

only display selected conditions on frequency panel

also change the number of significant figures on percentage scale

add CD19 channel and B-cells back

@gszep needs to optionally split on Foxp3-IgM if - exists in channel name, to Foxp3 and IgM as separate channels. For 390C fill IgM channel with NAN @mad3cientist we need to update all .fcs files (except 390C) with cleaned singlet live cells.

Hiding doms in JSServe v1.2.4 prevents OpenLayers from displaying tiles

One of the changes in this set of files SimonDanisch/Bonito.jl@v1.2.3...v1.2.4 prevents FlowAtlas.jl from displaying any tiles at all. For now I will restrict the version to be < 1.2.4

reset button

display percentage labels alongside gated populations

TagBot trigger issue

This issue is used to trigger TagBot; feel free to unsubscribe.

If you haven't already, you should update your TagBot.yml to include issue comment triggers.
Please see this post on Discourse for instructions and more details.

If you'd like for me to do this for you, comment TagBot fix on this issue.
I'll open a PR within a few hours, please be patient!

dynamic colorbar and gate by level

Optional batch normalisation to reference fcs files

The feature extractor currently uses a gaussian mixtures to shift and scale channels independently per sample to +/- 1. If reference files are present, use those to scale autoflouresnce to -1, leaving relative intensities in positive region alone.

• use Zoe's data for dev/testing