Save Segmentation Results Error

[ccnawrocki]

Hi there,
When I am running Baysor on CosMx data, it seems to work great for many of my fields of view (FOVs). However, with a select few, I get the following error:

(base) ccn22@MGB033458 hcc_tma2 % 
~/.julia/bin/baysor run FOV022.csv :cell_ID -g "target" -m 10 -x "x_local_px" -y "y_local_px" -s 70                            
[11:25:24] Info: Run R671c70508
[11:25:24] Info: (2024-12-12) Run Baysor v0.7.1
[11:25:24] Info: Loading data...
[11:25:27] Info: Loaded 615297 transcripts, 1087 genes.
[11:25:27] Info: Estimating noise level
[11:25:31] Info: Done
[11:25:35] Info: Clustering molecules...
Progress: 100%|███████████████████████████████████████████████████████████████████████████████████████████████████████████████| Time: 0:00:23
  Iteration:                  483
  Max. difference:            0.0048
  Fraction of probs changed:  0.235
[11:26:01] Info: Algorithm stopped after 483 iterations. Max. probability difference: 0.0048. Converged: true.
[11:26:01] Info: Done
[11:26:01] Info: Initializing algorithm. Scale: 70.0, scale std: 17.5, initial #components: 123058, #molecules: 615297.
[11:26:03] Info: Using the following additional information about molecules: [:confidence, :cluster, :prior_segmentation]
[11:26:03] Info: Using 3D coordinates
Progress: 100%|███████████████████████████████████████████████████████████████████████████████████████████████████████████████| Time: 0:01:34
  Iteration:        500
  Noise level, %:   1.91
  Num. components:  1847
[11:27:38] Info: Processing complete.
[11:27:39] Info: Estimating local colors
[11:28:09] Info: Estimating boundary polygons
Progress: 100%|███████████████████████████████████████████████████████████████████████████████████████████████████████████████| Time: 0:00:01
[11:28:13] Info: Saving results to segmentation.csv
ERROR: LoadError: MethodError: no method matching save_segmentation_results(::DataFrames.DataFrame, ::DataFrames.DataFrame, ::SparseArrays.SparseMatrixCSC{Int64, Int64}, ::OrderedCollections.OrderedDict{String, Dict{K, Matrix{Float64}} where K}, ::Baysor.Utils.OutputPaths; matrix_format::Symbol, gene_names::Vector{String}, polygon_format::String)

Closest candidates are:
  save_segmentation_results(::DataFrames.DataFrame, ::DataFrames.DataFrame, ::SparseArrays.SparseMatrixCSC{<:Real, Int64}, ::Union{Nothing, AbstractDict{TI, Matrix{TV}} where {TV<:Real, TI<:Union{Integer, String}}, AbstractDict{String, <:AbstractDict{TI, Matrix{TV}}} where {TV<:Real, TI<:Union{Integer, String}}}, ::Baysor.Utils.OutputPaths; polygon_format, gene_names, matrix_format)
   @ Baysor ~/.julia/packages/Baysor/7LPmz/src/data_loading/cli_wrappers.jl:224

Stacktrace:
 [1] run(coordinates::String, prior_segmentation::String; config::Baysor.Utils.RunOptions, x_column::Symbol, y_column::Symbol, z_column::Symbol, gene_column::Symbol, min_molecules_per_cell::Int64, scale::Float64, scale_std::String, n_clusters::Int64, prior_segmentation_confidence::Float64, output::String, plot::Bool, polygon_format::String, count_matrix_format::String)
   @ Baysor.CommandLine ~/.julia/scratchspaces/cc9f9468-1fbe-11e9-0acf-e9460511877c/sysimg/libbaysor.dylib:-1
in expression starting at /Users/ccn22/.julia/bin/baysor:15

Everything seems to run fine, but there is a problem with saving.

I am using Baysor version 0.7.1 on MacOS 15.1.1
Julia version 1.10.7
I have an M3 MacBook Pro

Any help would be much appreciated.

[ccnawrocki]

Update: I have identified the problem. I read the transcripts (tx) files into R for a few of the select FOVs that did not work. After grouping by z position and counting, I get a table like this for all of the FOVs:

       z      n
   <int>  <int>
 1    -1      2
 2     0   8407
 3     1  16210
 4     2  28503
 5     3  52166
 6     4  89855
 7     5 117445
 8     6 134940
 9     7 167761
10     8      8

In each FOV, one of the z positions has very few tx associated with it. In this example, after removing the two tx that had a z position of -1, Baysor worked. So, in the general case, if I simply exclude all tx that occur in a certain z plane that contains less than say 5 tx, then I am in business.

Is there an option or flag that I am missing that would resolve this problem?

[VPetukhov]

Thank you for debugging it, @ccnawrocki ! That's the first time I see negative z. That might be the problem. I will test it more for the next release. Currently there are no flags to deal with it, unless you want to convert it all to 2D (--config.data.force-2d=true)