| Title: | Classify Digital PCR Droplets by Fitting Fluorescence Populations |
|---|---|
| Description: | Estimates DNA target concentration by classifying digital PCR (polymerase chain reaction) droplets as positive, negative, or rain, using Expectation-Maximization Clustering. The fitting is accomplished using the 'EMMIXskew' R package (v. 1.0.3) by Kui Wang, Angus Ng, and Geoff McLachlan (2018) as based on their paper "Multivariate Skew t Mixture Models: Applications to Fluorescence-Activated Cell Sorting Data" <doi:10.1109/DICTA.2009.88>. |
| Authors: | Joyce Emlyn Guiao [aut, cre] |
| Maintainer: | Joyce Emlyn Guiao <[email protected]> |
| License: | GPL (>= 3) |
| Version: | 0.1.1.1 |
| Built: | 2024-10-14 03:46:08 UTC |
| Source: | https://github.com/zeroh729/poppcr |
Mean target copies per partition (lambda) is derived using Poisson distribution as lambda = -ln(nneg / ntot). Target copies in sample is then calculated as conc = lambda * volSamp/(volDrp * 1000).
calculateConc(nneg, ntotal, volSamp, volDrp)calculateConc(nneg, ntotal, volSamp, volDrp)
nneg |
numeric, negative droplet count |
ntotal |
numeric, total droplet count |
volSamp |
numeric, sample volume in microliter |
volDrp |
numeric, droplet (or partition) volume in nanoliter |
Returns a list with 2 named items lambda and conc
lambda - numeric, vector of mean target copies per partition (lambda) and its lower and upper 95% confidence interval
conc - numeric, vector of target copies in sample (based on the given sample volume (volSamp) and droplet volume (volDrp)) and its lower and upper 95% confidence interval
estimates <- calculateConc(5000, 20000, volSamp = 20, volDrp = 0.85) estimates # Output: # $lambda # lambda lower upper # 1.386294 1.362289 1.410299 # # $conc # conc lower upper # 32618.69 32053.87 33183.51estimates <- calculateConc(5000, 20000, volSamp = 20, volDrp = 0.85) estimates # Output: # $lambda # lambda lower upper # 1.386294 1.362289 1.410299 # # $conc # conc lower upper # 32618.69 32053.87 33183.51
Estimates target concentration by counting positive droplets with Poisson correction. Positive, negative, and rain populations are fitted using EM. Droplets are then classified using Maximum A Posteriori rule
popPCR( x, dist, volSamp = 20, volDrp = 0.85, maxComponents = Inf, negProbThres = 1e-07, useOnlyNegProbThres = FALSE )popPCR( x, dist, volSamp = 20, volDrp = 0.85, maxComponents = Inf, negProbThres = 1e-07, useOnlyNegProbThres = FALSE )
x |
numeric, vector of droplet fluorescence amplitude |
dist |
character, distribution of the mixture models ("normal", "skewed-normal", "t", "skewed-t") |
volSamp |
numeric, sample volume in microliter |
volDrp |
numeric, droplet (or partition) volume in nanoliter |
maxComponents |
numeric, maximum number of components (e.g. populations) |
negProbThres |
numeric, if only one population was detected, then its assumed as a negative population. Droplets will be classified as positive if its probability given the population < negProbThres. |
useOnlyNegProbThres |
logical, if TRUE, then droplets will be classified as positive if its probability given the leftmost population < negProbThres. Default is FALSE, i.e. classification is done by Maximum A Posteriori rule. |
Returns a result.popPCR S4 class object with attributes
classification - character, vector of droplet classification
dist - character, user-specified parameter for the mixture model
dropletCount - list, droplet classification count
em - list, returned value of EMMIXskew's EmSkew()
estConc - list, estimated target concentration as lambda and sample concentration (with 95% CI)
G - numeric, number of components fitted
memberProb - list, component membership probability of all droplets
library(popPCR) # Plot histograms of available data hist(x_onePop, breaks = 100) hist(x_twoPop, breaks = 100) hist(x_multiPop, breaks = 100) # ---- Mixture model fitting ---- # # Example 1. One population sample result <- popPCR(x_onePop, dist = "t") printSummaryConc(result) # Output: # Populations detected : 1 # Total droplets : 8000 # Positive : 1 (0.01%) # Negative : 7999 (99.99%) # # Target copies in sample : 2.9414 ( 95% CI: [ -2.8237 , 8.7064 ] ) # Mean target copies per partition : 1e-04 ( 95% CI: [ -1e-04 , 4e-04 ] ) printSummaryFit(result) # Output: # Results of fitting a 1-component t mixture model # # Negative Population # Mix prop. : 1 # Mu : 1024.1614 # Sigma : 35253.1747 # Dof : 2.005 # (Option) increase negProbThres to classify negative droplets more strictly result <- popPCR(x_onePop, dist = "t", negProbThres = 1e-4) printSummaryConc(result) # Output: # Populations detected : 1 # Total droplets : 8000 # Positive : 691 (8.64%) # Negative : 7309 (91.36%) # # Target copies in sample : 2125.5312 ( 95% CI: [ 1966.9936 , 2284.0688 ] ) # Mean target copies per partition : 0.0903 ( 95% CI: [ 0.0836 , 0.0971 ] ) # Example 2. Two population sample result <- popPCR(x_twoPop, dist = "t") printSummaryConc(result) # Output: # Populations detected : 2 # Total droplets : 10254 # Positive : 8693 (84.78%) # Negative : 1561 (15.22%) # # Target copies in sample : 44290.3819 ( 95% CI: [ 43215.6408 , 45365.1231 ] ) # Mean target copies per partition : 1.8823 ( 95% CI: [ 1.8367 , 1.928 ] ) printSummaryFit(result) # Output: # Results of fitting a 2-component t mixture model # # Negative Population # Mix prop. : 0.1522 # Mu : 2136.7435 # Sigma : 4126.8357 # Dof : 12.3562 # # Positive Population # Mix prop. : 0.8478 # Mu : 7580.1275 # Sigma : 42621.1894 # Dof : 2.415 # Example 3. Multiple population sample result <- popPCR(x_multiPop, dist = "t", maxComponents = 4) printSummaryConc(result) # Output: # Populations detected : 4 # Total droplets : 1814 # Positive : 44 (2.43%) # Negative : 1252 (69.02%) # Rain (1) : 258 (14.22%) # Rain (2) : 260 (14.33%) # # Target copies in sample : 8724.5195 ( 95% CI: [ 7999.0578 , 9449.9812 ] ) # Mean target copies per partition : 0.3708 ( 95% CI: [ 0.34 , 0.4016 ] ) # In the output above, we see 2 rain populations! Let's examine its plot. plot(stats::density(x_multiPop)) # We can see that Rain (1) is very close to the Negative population. # Let's include droplets in Rain (1) in the negative droplet count. nNegative <- result@dropletCount$neg + result@dropletCount$rain1 nTotal <- result@dropletCount$total # Re-estimate concentration as follows newEstimates <- calculateConc(nNegative, nTotal, volSamp = 20, volDrp = 0.85) newEstimates # Output: # $lambda # lambda lower upper # 0.1834247 0.1627763 0.2040731 # # $conc # conc lower upper # 4315.875 3830.031 4801.719library(popPCR) # Plot histograms of available data hist(x_onePop, breaks = 100) hist(x_twoPop, breaks = 100) hist(x_multiPop, breaks = 100) # ---- Mixture model fitting ---- # # Example 1. One population sample result <- popPCR(x_onePop, dist = "t") printSummaryConc(result) # Output: # Populations detected : 1 # Total droplets : 8000 # Positive : 1 (0.01%) # Negative : 7999 (99.99%) # # Target copies in sample : 2.9414 ( 95% CI: [ -2.8237 , 8.7064 ] ) # Mean target copies per partition : 1e-04 ( 95% CI: [ -1e-04 , 4e-04 ] ) printSummaryFit(result) # Output: # Results of fitting a 1-component t mixture model # # Negative Population # Mix prop. : 1 # Mu : 1024.1614 # Sigma : 35253.1747 # Dof : 2.005 # (Option) increase negProbThres to classify negative droplets more strictly result <- popPCR(x_onePop, dist = "t", negProbThres = 1e-4) printSummaryConc(result) # Output: # Populations detected : 1 # Total droplets : 8000 # Positive : 691 (8.64%) # Negative : 7309 (91.36%) # # Target copies in sample : 2125.5312 ( 95% CI: [ 1966.9936 , 2284.0688 ] ) # Mean target copies per partition : 0.0903 ( 95% CI: [ 0.0836 , 0.0971 ] ) # Example 2. Two population sample result <- popPCR(x_twoPop, dist = "t") printSummaryConc(result) # Output: # Populations detected : 2 # Total droplets : 10254 # Positive : 8693 (84.78%) # Negative : 1561 (15.22%) # # Target copies in sample : 44290.3819 ( 95% CI: [ 43215.6408 , 45365.1231 ] ) # Mean target copies per partition : 1.8823 ( 95% CI: [ 1.8367 , 1.928 ] ) printSummaryFit(result) # Output: # Results of fitting a 2-component t mixture model # # Negative Population # Mix prop. : 0.1522 # Mu : 2136.7435 # Sigma : 4126.8357 # Dof : 12.3562 # # Positive Population # Mix prop. : 0.8478 # Mu : 7580.1275 # Sigma : 42621.1894 # Dof : 2.415 # Example 3. Multiple population sample result <- popPCR(x_multiPop, dist = "t", maxComponents = 4) printSummaryConc(result) # Output: # Populations detected : 4 # Total droplets : 1814 # Positive : 44 (2.43%) # Negative : 1252 (69.02%) # Rain (1) : 258 (14.22%) # Rain (2) : 260 (14.33%) # # Target copies in sample : 8724.5195 ( 95% CI: [ 7999.0578 , 9449.9812 ] ) # Mean target copies per partition : 0.3708 ( 95% CI: [ 0.34 , 0.4016 ] ) # In the output above, we see 2 rain populations! Let's examine its plot. plot(stats::density(x_multiPop)) # We can see that Rain (1) is very close to the Negative population. # Let's include droplets in Rain (1) in the negative droplet count. nNegative <- result@dropletCount$neg + result@dropletCount$rain1 nTotal <- result@dropletCount$total # Re-estimate concentration as follows newEstimates <- calculateConc(nNegative, nTotal, volSamp = 20, volDrp = 0.85) newEstimates # Output: # $lambda # lambda lower upper # 0.1834247 0.1627763 0.2040731 # # $conc # conc lower upper # 4315.875 3830.031 4801.719
Summarizes the number of populations detected, total droplets, and number of classified positive, negative, and rain droplets. Also calculates the target copies in sample and the mean target copies per partition (lambda).
printSummaryConc(result.popPCR)printSummaryConc(result.popPCR)
result.popPCR |
returned value of popPCR() |
result <- popPCR(x_twoPop, dist = "t") printSummaryConc(result) # Output: # Populations detected : 2 # Total droplets : 10254 # Positive : 8693 (84.78%) # Negative : 1561 (15.22%) # # Target copies in sample : 44290.3819 ( 95% CI: [ 43215.6408 , 45365.1231 ] ) # Mean target copies per partition : 1.8823 ( 95% CI: [ 1.8367 , 1.928 ] )result <- popPCR(x_twoPop, dist = "t") printSummaryConc(result) # Output: # Populations detected : 2 # Total droplets : 10254 # Positive : 8693 (84.78%) # Negative : 1561 (15.22%) # # Target copies in sample : 44290.3819 ( 95% CI: [ 43215.6408 , 45365.1231 ] ) # Mean target copies per partition : 1.8823 ( 95% CI: [ 1.8367 , 1.928 ] )
Summarizes the number of populations fitted and their estimate distribution parameters. If only 1 population was detected, then it is assumed and is identified to be a negative population. If 2 populations were detected, then the leftmost is identified as the Negative Population and the rightmost is the Positive Population. If 3 or more populations were detected, then the populations between the leftmost and the rightmost will be considered as Rain Populations; which are numbered to make it identifiable in case of multiple Rain Populations (i.e. Rain (1) and Rain (2)).
printSummaryFit(result.popPCR)printSummaryFit(result.popPCR)
result.popPCR |
returned value of popPCR() |
result <- popPCR(x_twoPop, dist = "t") printSummaryFit(result) # Output: # Results of fitting a 2-component t mixture model # # Negative Population # Mix prop. : 0.1522 # Mu : 2136.7435 # Sigma : 4126.8357 # Dof : 12.3562 # # Positive Population # Mix prop. : 0.8478 # Mu : 7580.1275 # Sigma : 42621.1894 # Dof : 2.415 result <- popPCR(x_multiPop, dist = "t", maxComponents = 4) printSummaryFit(result) # Output: # Results of fitting a 4-component t mixture model # # Negative Population # Mix prop. : 0.6896 # Mu : 1452.1416 # Sigma : 12526.8931 # Dof : 21.3612 # # Rain (1) Population # Mix prop. : 0.142 # Mu : 2142.1118 # Sigma : 10762.5474 # Dof : 186.2947 # # Rain (2) Population # Mix prop. : 0.1457 # Mu : 5119.0039 # Sigma : 334959.2499 # Dof : 2.3626 # # Positive Population # Mix prop. : 0.0227 # Mu : 8505.9682 # Sigma : 192858.9044 # Dof : 149.8677result <- popPCR(x_twoPop, dist = "t") printSummaryFit(result) # Output: # Results of fitting a 2-component t mixture model # # Negative Population # Mix prop. : 0.1522 # Mu : 2136.7435 # Sigma : 4126.8357 # Dof : 12.3562 # # Positive Population # Mix prop. : 0.8478 # Mu : 7580.1275 # Sigma : 42621.1894 # Dof : 2.415 result <- popPCR(x_multiPop, dist = "t", maxComponents = 4) printSummaryFit(result) # Output: # Results of fitting a 4-component t mixture model # # Negative Population # Mix prop. : 0.6896 # Mu : 1452.1416 # Sigma : 12526.8931 # Dof : 21.3612 # # Rain (1) Population # Mix prop. : 0.142 # Mu : 2142.1118 # Sigma : 10762.5474 # Dof : 186.2947 # # Rain (2) Population # Mix prop. : 0.1457 # Mu : 5119.0039 # Sigma : 334959.2499 # Dof : 2.3626 # # Positive Population # Mix prop. : 0.0227 # Mu : 8505.9682 # Sigma : 192858.9044 # Dof : 149.8677
The reaction with ID 373 in the Dataset.zip provided in the repository from Lievens et. al. (2017)
x_multiPopx_multiPop
A numeric vector with 1814 droplet fluorescence
https://github.com/Gromgorgel/ddPCR/blob/master/Dataset.zip
Simulated dataset with very high presence of rain and true mean copies per partition of 0.1.
x_onePopx_onePop
A numeric vector with 8000 droplet fluorescence
The reaction with ID 9 in the Dataset.zip provided in the repository from Lievens et. al. (2017)
x_twoPopx_twoPop
A numeric vector with 10254 droplet fluorescence
https://github.com/Gromgorgel/ddPCR/blob/master/Dataset.zip