Quality-control thresholds

Builds the thresholds that load_rcc() and normalise() use to flag samples.

Usage

nacho_thresholds(
  instrument = c("max", "flex", "pro", "sprint"),
  preset = c("nsolver", "legacy"),
  haemolysis = FALSE
)

Arguments

instrument

The nCounter instrument: "max", "flex", "pro" or "sprint".

preset

"nsolver" or "legacy".

haemolysis

If TRUE, flag samples of a miRNA panel whose miR-451a to miR-23a-3p log2 ratio is above 7, a sign of haemolysis in plasma and serum. load_rcc() does not take it, so pass the thresholds to normalise(outliers_thresholds = ).

Value

A named list: preset, instrument, then one element per metric.

Details

"nsolver" follows the current Bruker nCounter knowledge base (https://brukerspatialbiology.com/support/, "nCounter QC flags"): binding density 0.05 to 2.25 on MAX, FLEX and PRO and 0.1 to 1.8 on SPRINT, field of view at least 75 %, positive control linearity at least 0.95 (POS_A to POS_E, with log2(count + 1)), limit of detection at least 2, positive factor 0.3 to 3 and housekeeping factor 0.1 to 10. It also asks for at least three housekeeping genes above background, as Bruker asks (Housekeeping_detected).

"legacy" reproduces NACHO 2: binding density 0.1 to 2.25 on every instrument, positive control linearity with POS_F, positive factor 1/4 to 4 and housekeeping factor 1/11 to 11. With background = "geo", background_mode = "subtract", it gives back the NACHO 2 outlier calls.

The preset also chooses how negative probes are excluded and how the positive control linearity is computed, so changing it normalises again.

For miRNA panels, Ligation_order (the three ligation positive controls in decreasing order), Ligation_R2 (their log2 counts on a line) and Ligation_NEG (the largest ligation negative minus the detection limit) are NACHO's own definitions, since Bruker publishes no threshold. "legacy" never flags them.

Whole-blood RNA raises LIG_NEG_A, and to a lesser extent LIG_NEG_B, together with red blood cell miRNAs such as miR-451a. The ligation negatives of such samples are above the detection limit, so most samples of a whole-blood study fail Ligation_NEG. This pattern occurs in independent PAXgene whole-blood studies on GEO (GSE303669 and GSE317496), and not in plasma, serum or extracellular vesicle studies. When more than half of the samples fail Ligation_NEG, load_rcc() warns about this pattern. For whole-blood samples, set a higher upper bound for Ligation_NEG and pass the thresholds to normalise(outliers_thresholds = ). Do not remove these samples with exclude_outliers() because of this flag only.

Each element can be changed afterwards; two numbers are a lower and an upper bound, and one number is a lower bound. Use -Inf or Inf for an open bound.

Examples

nacho_thresholds()
#> $preset
#> [1] "nsolver"
#> 
#> $instrument
#> [1] "max"
#> 
#> $BD
#> [1] 0.05 2.25
#> 
#> $FoV
#> [1] 75
#> 
#> $PCL
#> [1] 0.95
#> 
#> $LoD
#> [1] 2
#> 
#> $Positive_factor
#> [1] 0.3 3.0
#> 
#> $House_factor
#> [1]  0.1 10.0
#> 
#> $Housekeeping_detected
#> [1] 3
#> 
#> $Ligation_order
#> [1] 1
#> 
#> $Ligation_R2
#> [1] 0.95
#> 
#> $Ligation_NEG
#> [1] -Inf    0
#> 
#> $Haemolysis
#> [1] -Inf  Inf
#> 
nacho_thresholds("sprint")
#> $preset
#> [1] "nsolver"
#> 
#> $instrument
#> [1] "sprint"
#> 
#> $BD
#> [1] 0.1 1.8
#> 
#> $FoV
#> [1] 75
#> 
#> $PCL
#> [1] 0.95
#> 
#> $LoD
#> [1] 2
#> 
#> $Positive_factor
#> [1] 0.3 3.0
#> 
#> $House_factor
#> [1]  0.1 10.0
#> 
#> $Housekeeping_detected
#> [1] 3
#> 
#> $Ligation_order
#> [1] 1
#> 
#> $Ligation_R2
#> [1] 0.95
#> 
#> $Ligation_NEG
#> [1] -Inf    0
#> 
#> $Haemolysis
#> [1] -Inf  Inf
#> 
nacho_thresholds(preset = "legacy")
#> $preset
#> [1] "legacy"
#> 
#> $instrument
#> [1] "max"
#> 
#> $BD
#> [1] 0.10 2.25
#> 
#> $FoV
#> [1] 75
#> 
#> $PCL
#> [1] 0.95
#> 
#> $LoD
#> [1] 2
#> 
#> $Positive_factor
#> [1] 0.25 4.00
#> 
#> $House_factor
#> [1]  0.09090909 11.00000000
#> 
#> $Housekeeping_detected
#> [1] 0
#> 
#> $Ligation_order
#> [1] 0
#> 
#> $Ligation_R2
#> [1] 0
#> 
#> $Ligation_NEG
#> [1] -Inf  Inf
#> 
#> $Haemolysis
#> [1] -Inf  Inf
#>