DataPlates.jl

Julia Package for working with data on SBS or microtiter (MTP) plates.

This is a rewrite of MTP.jl which was never registered.

Similar use-cases in R: https://github.com/ropensci/plater, https://github.com/jpquast/ggplate

Draft

Below are implementation thoughts.

Not much done yet:

• struct DataPlate
• function to generate wells: A01, A02, ....
• Only depend on Tables rather than DataFrames
• Document constructor and conversion to DataFrame

Structs

struct DataPlate
    name::String
    barcode::String
    geometry::Int
	quadrant_pattern::Vector{Int}
    values::Vector{WellValues}
end

Assert that all value-vectors have length == geometry.

struct WellValues{T} T
    name::String ## the "column name"
    values::Vector{T} ## Well values are always stored column-wise
end

Well values are always stored column-wise. Re-sort on well to get them row-wise
T can be numeric or String.
T can itself be a vector, so this can also hold time-series as reader-curves.
But we probably want to keep this use case separate.

struct SparseWellValue{T}
  well_name::String
   value::T
end

Constructors

• DataPlate("Plate1"): 96 well plate with name "Plate1", same as barcode one value: "well_96" with values "A01", "B01", ....
• DataPlate("Plate1", "UP0001234"): 96 well plate with name: "Plate1", barcode: "UP0001234" and one value with one value: "well_96" with values "A01", "B01", ....
• Do not allow "well", "platename", "barcode" as a column names. These are reserved.
• 6-well plates only have a single "quadrant" [1]
• values!(p::DataPlate, colname::String, values::Vector{T}): add a column
• DataPlate(;col1 = values1, col2 = values2, ...) plate with columns

Methods

• Fill a plate with sparse well values.
• DataFrame: convert to tabular form
• print: print in plate form
• write: write a vector of plates as excel workbook
• read: read vector of plates from excel workbook
• format well names
• compute quadrant in different Q patterns:
  • Q pattern: list quadrant numbers in "math order"
  • Normal "Z" pattern is then: 2,1,3,4
  • "Math order" is "C" pattern
• merge 96 well plates to 384 using quadrant-pattern

Recipes

• plot similar to ggplate https://github.com/jpquast/ggplate

Input format

Data is read in from a "plate format" like the one used in plater:

The "Name" is used a column name for the values in the "wells".
If read in from a spread-sheet workbook, the sheet name is used as plate name, and the file name is used as plateset name.

When the data is converted to DataFrame is looks like this:

If the input data is sparse (not all wells have values), the DataFrame can be sparse as well (only wells with values) or full depending on a kw parameter.

Design considerations

well order

• Do we want a fixed well order?

It makes everything easier, if we just decide on an order.
If we sort by column as default (A01, B01, ...) it is easy to resort to row-wise (A01, A02, ...)

well names

• Do we want well names as "A01" or "A1"?

The "A01" is nice as it sorts in data frames and spread sheets.
However, it does not scale to 1536, where we have wells: A01, ..., Z48, AA01, ..., AF48

We will consider 1536, when we get there.

The advantage of A01 wins for now.

Examples

This is going into docs.

julia> wells(6)
6-element Vector{String}:
"A01"
"B01"
"A02"
"B02"
"A03"
"B03"

julia> sort(wells(6))
6-element Vector{String}:
"A01"
"A02"
"A03"
"B01"
"B02"
"B03"

julia> DataFrame(DataPlate("Plate1",6))
6×4 DataFrame
Row │ platename barcode well well006
│ String String String String
─────┼─────────────────────────────────────
1 │ Plate1 Plate1 A01 A01
2 │ Plate1 Plate1 B01 B01
3 │ Plate1 Plate1 A02 A02
4 │ Plate1 Plate1 B02 B02
5 │ Plate1 Plate1 A03 A03
6 │ Plate1 Plate1 B03 B03

StructEquality gives us:

julia> DataPlate("Plate1") == DataPlate("Plate1", "Plate1")
julia> DataPlate("Plate1") == DataPlate("Plate1", "Plate1", 96)

TODO:
julia> DataFrame(DataPlate("Plate1",6; activity = [1,2,3,4,5,6]))

row, col should check that input has form r"\D+\d+"

Function for generating unique barcodes.

• Base uniqueness on prefix and epoc timestamp.
• Limit total number of characters, as barcode printers and readers have limitations.
• generate set of barcods in batches

Make barcodes funtion (prefix, width; count=1, start=missing, sep="")
Compose by prefix, start, sep, index
Index Counts Up from 1.
For index missing use last n digits of epoch.
N= width - length(prefix) - length(sep) - digits(count)