Prepares real and simulated matrices for DISSECT fraction estimation by applying the same preprocessing order used in the original workflow: transformation, variance filtering, deduplication, normalisation, gene intersection, and size balancing.
Usage
dissect_process(
bulk,
reference = NULL,
sim_data = NULL,
test_dataset_type = "bulk",
duplicated = "first",
normalize_simulated = "cpm",
normalize_test = "cpm",
var_cutoff = 0.1,
test_in_mix = 1
)Arguments
- bulk
Numeric matrix with genes in rows and samples in columns.
- reference
Numeric matrix with genes in rows and cell types in columns, or `NULL`. This is a convenience path and is only used when `sim_data` is not supplied.
- sim_data
A list returned by [dissect_simulate()], or `NULL`.
- test_dataset_type
Character scalar. Either `"bulk"` or `"microarray"`.
- duplicated
Character scalar. How duplicated gene names should be resolved. One of `"first"`, `"sum"`, or `"mean"`.
- normalize_simulated
Character scalar or `NULL`. Currently `"cpm"` or `NULL`.
- normalize_test
Character scalar or `NULL`. Currently `"cpm"` or `NULL`.
- var_cutoff
Numeric scalar or `NULL`. Variance threshold applied to the bulk input before transposition.
- test_in_mix
Integer scalar. Number of real samples used in the online mixing step.
Value
A named list with components:
- X_real_train
Real training matrix with samples in rows and genes in columns.
- X_sim
Simulated matrix with samples in rows and genes in columns.
- y_sim
Simulated proportions with samples in rows and cell types in columns.
- X_real_test
Real test matrix with samples in rows and genes in columns.
- sample_names
Character vector of sample names.
- celltypes
Character vector of cell-type names.
- genes
Character vector of common genes used for modelling.
- reference
Reference matrix if supplied, otherwise `NULL`.
- sim_data
Simulation object if supplied, otherwise `NULL`.
Details
The intended DISSECT workflow uses `sim_data` generated from [dissect_simulate()]. A direct `reference` matrix is supported as a convenience interface for proportion estimation.
References
Khatri, R., Machart, P., & Bonn, S. (2024). DISSECT: deep semi-supervised consistency regularization for accurate cell type fraction and gene expression estimation. Genome Biology, 25(1), 112.
Original DISSECT software repository: https://github.com/imsb-uke/DISSECT