scran::ChoosePseudoCount Class Reference

Choose a pseudo-count for log-transformation. More...

#include <ChoosePseudoCount.hpp>

Classes
struct	Defaults
	Default parameters. More...

Public Member Functions
ChoosePseudoCount &	set_quantile (double q=Defaults::quantile)
ChoosePseudoCount &	set_max_bias (double b=Defaults::max_bias)
ChoosePseudoCount &	set_min_value (double v=Defaults::min_value)
double	run (size_t n, const double *size_factors, double *buffer) const
double	run (size_t n, const double *size_factors) const

Detailed Description

Choose a pseudo-count for log-transformation.

This class chooses a pseudo-count for log-transformation (see LogNormCounts::set_pseudo_count()) that aims to control the log-transform bias. Specifically, the log-transform can introduce spurious differences in the expected log-normalized expression between cells with very different size factors. This bias can be mitigated by increasing the pseudo-count, which effectively shrinks all log-expression values towards the zero-expression baseline. The increased shrinkage is strongest at low counts where the log-transform bias is most pronounced, while the transformation of large counts is mostly unaffected.

In practice, the log-transformation bias is modest in datasets where there are stronger sources of variation. When observed, it manifests as a library size-dependent trend in the log-normalized expression values. This is difficult to regress out without also removing biology that is associated with, e.g., total RNA content; rather, a simpler solution is to increase the pseudo-count to suppress the bias.

No centering is performed by this function, so if centering is required, the size factors should be used in CenterSizeFactors first.

See also

Lun ATL (2018). Overcoming systematic errors caused by log-transformation of normalized single-cell RNA sequencing data biorXiv doi:10.1101/404962

Member Function Documentation

set_quantile()

ChoosePseudoCount & scran::ChoosePseudoCount::set_quantile ( double  q = Defaults::quantile )

inline

Parameters

q	Quantile to use for finding the smallest/largest size factors. Setting this to zero will use the observed minimum and maximum, though this is usually too extreme in practice. The default is to take the 5th and 95th percentile, yielding a range that is still representative of most cells.

Returns

Reference to this ChoosePseudoCount class.

set_max_bias()

ChoosePseudoCount & scran::ChoosePseudoCount::set_max_bias ( double  b = Defaults::max_bias )

inline

Parameters

b	Acceptable upper bound on the log-transformation bias.

Returns

Reference to this ChoosePseudoCount class.

set_min_value()

ChoosePseudoCount & scran::ChoosePseudoCount::set_min_value ( double  v = Defaults::min_value )

inline

Parameters

v	Minimum value for the pseudo-count returned by run(). Defaults to 1 to stabilize near-zero normalized expression values, otherwise these manifest as avoid large negative values.

Returns

Reference to this ChoosePseudoCount class.

run() [1/2]

double scran::ChoosePseudoCount::run ( size_t  n, const double *  size_factors, double *  buffer  ) const

inline

Parameters

	n	Number of size factors.
[in]	size_factors	Pointer to an array of size factors of length n. Values should be positive, and all non-positive values are ignored.
	buffer	Pointer to an array of length n, to be used as a workspace.

Returns

The suggested pseudo-count to control the log-transformation-induced bias below the specified threshold.

run() [2/2]

double scran::ChoosePseudoCount::run ( size_t  n, const double *  size_factors  ) const

inline

Parameters

	n	Number of size factors.
[in]	size_factors	Pointer to an array of size factors of length n. Values should be positive, and all non-positive values are ignored.

Returns

The suggested pseudo-count to control the log-transformation-induced bias below the specified threshold.

---

The documentation for this class was generated from the following file:

• scran/normalization/ChoosePseudoCount.hpp