ENH: Improvement to trackClonotypes

[EugeneRumynskiy]

🚀 Feature

Improvement to trackClonotypes(...)

Motivation

Pitch

Alternatives

Additional context

Currently, we have an initial tech spec for the function, that allows to start coding while I’m preparing the full Tech

Spec:

• works on the standard data-metadata list (using the “Timepoint” field as a variable to order samples on the time line)
• works on any number of samples
• works on any columns (see the .col argument in trackClonotypes, feel free to copy the functionality from there - we will integrate your code inside trackClonotypes anyways)
• doesn’t take into accounts replicas (we will add later if needed)

Internals:

• subsets the data (using the Timepoint field to distinguish days for the subsequent clustering. Throw an error if no Timepoint field)
• applies PCA
• clusters trajectories using the best cluster determined automatically
  outputs the list with:
• $pca - the PCA object
• $data - the data frame with sequences, columns and the Cluster columns before the Samples with cluster indices, so people can easily use it later
• $traj - trajectories
• class immunr_trajectories
• visualisations can be easily added later

[vadimnazarov]

Additionally, for the documentation:

• add links everywhere into "See Also"
• update the tutorials inside functions' references

[vadimnazarov]

ToDo for @vadimnazarov :

• edge cases
• progress bars

[vadimnazarov]

• trackClonotypes2 changes:
  • make sure there are two columns presented: "Timepoint" (for time points) and "Origin" (for patients)
  • Edge cases:
    • N: <0
    • metadata should contain the "Timepoints" column
  • Put in the doc that don't pass a large N if you don't have enough RAM

[vadimnazarov]

progress bars – it's OK for now

[vadimnazarov]

test it on the COVID-19 tutorial data before integrating it to the trackClonotypes