/doubletQC

Quality check of doublet detection methods in single-cell RNA-seq data

Primary LanguagePythonMIT LicenseMIT

doubletQC

Benchmarking of doublet detection methods in single-cell RNA-seq data

Method list

  • Scrublet

    Model parameter:

    • r: default 2. The ratio of simulated cells
    • K: sqrt(n_input_cell) / 2. The number of neighbours in KNN

    Threshold parameter:

    • doublet score: threshold to set. Can be optimised by Bayesian Gaussian mixture model, added in the wrapper function here.

  • DoubletFinder

    Model parameter:

    • pN: default 0.25. The ratio of simulated cells r = pN / (1 - pN). This parameter pN has been shown resistant in the paper
    • pK: propotion of neighbours in KNN. Can be optimised with build-in function

    Threshold parameter:

    • nExp: number of expected to heterotypic doublets for threshold
    • pANN: fraction of simulated doublet neighbours, threshold with nExp

  • DoubletDecon

Wrapper functions

See wrapper functions in the bin folder

Example