| Title: | Estimate Insect Survival from Fluorescence Data |
|---|---|
| Description: | Use spectrophotometry measurements performed on insects as a way to infer pathogens virulence. Insect movements cause fluctuations in fluorescence signal, and functions are provided to estimate when the insect has died as the moment when variance in autofluorescence signal drops to zero. The package provides functions to obtain this estimate together with functions to import spectrophotometry data from a Biotek microplate reader. Details of the method are given in Parthuisot et al. (2018) <doi:10.1101/297929>. |
| Authors: | Jean-Baptiste Ferdy <[email protected]> |
| Maintainer: | Jean-Baptiste Ferdy <[email protected]> |
| License: | GPL-3 |
| Version: | 1.0.0 |
| Built: | 2024-09-19 03:42:53 UTC |
| Source: | https://github.com/cran/fluoSurv |
Estimation of time to death
estimate.LT(y, t, threshold.value = NULL, offset.dead = NULL, offset.alive = NULL, verbose = F, ndeps = 0.01)estimate.LT(y, t, threshold.value = NULL, offset.dead = NULL, offset.alive = NULL, verbose = F, ndeps = 0.01)
y |
The signal to analize |
t |
The time values |
threshold.value |
Detection threshold below which signal is considered as pure noise |
offset.dead |
Offset derived from variance in signal in dead insects |
offset.alive |
Offset derived from variance in signal in living insects |
verbose |
If true additionnal informations on computations are displayed |
ndeps |
Step size used in optim to estimate parameters |
The model adjusted in this procedure assumes that random variance in signal drops after insect death, because insect has ceased to move. The time value at which this drop occurs can therefore be used as an estimate of lethal time. First guess of variance estimates can be provided as offset.dead (for dead insects) or offset.alive (for living insects).
A vector with TL estimate for the given sample, corresponding log-likelihood and number of values used in the computation.
##Loading data data("galleria") ## dataset may contain NA if microplate reader has been stoped before the programmed ## end of the experiment d <- subset(galleria,!is.na(value)) l <- lapply(split(d,d$well),extract.well) data(setup) setup <- setup[match(setup$well,names(l)),] ## Computes rough estimates of variance in autofluorescence signal for dead and living insects ## These values serve as initial guess to fit the model. offset.alive <- estimate.offset(l,"1_330_405",min.t=1,max.t=5) # all insects are assumed to be alive during the # first five hours that follow injection offset.dead <- estimate.offset(l[which(setup$dead==1)],"1_330_405",min.t=72-5) # insects that were dead at the end of the experiment are # assumed to have died earlier than 5 hours before then # end of the experiment ## LT estimation or for a single well ## Check out well D9, to see what happens when an insect stayed alive. well <- "A3" plot(log(value_1_485_535,10)~t_1_485_535,type="l",col="green",ylim=c(2,5),data=l[[well]]) points(log(value_1_330_405,10)~t_1_330_405,type="l",col="gray",data=l[[well]]) # Rough estimate obtained using no offsets est1 <- with(l[[well]], estimate.LT(value_1_330_405,t_1_330_405,threshold.value=3)) abline(v=est1[["LT"]],lty=2,col="red") # Much better estimate obtained using offset for dead insects est2 <- with(l[[well]], estimate.LT(value_1_330_405,t_1_330_405, offset.dead = offset.dead,threshold.value=3)) abline(v=est2["LT"],lty=3,col="red") # Using offset.alive does not change anything to the estimate for well A3 # It may help for insect that have a larger variance in signal than others even after death est3 <- with(l[[well]], estimate.LT(value_1_330_405,t_1_330_405, offset.dead = offset.dead, offset.alive = offset.alive,threshold.value=3)) abline(v=est3["LT"],col="red") # Detection of significant GFP fluorescence (i.e. log fluorescence exceed # by 5% the maximum value observed during the first five hours) with(l[[well]],abline(v=when.threshold(t_1_485_535,log(value_1_485_535,10), threshold=0.1),col="green")) ##LT estimation for all wells if(FALSE) { #example takes time! Set to TRUE if you want to run it res <- sapply(l,function(x) estimate.LT(x$value_1_330_405,x$t_1_330_405, threshold.value=2,offset.dead=offset.dead, offset.alive=offset.alive)) res <- as.data.frame(t(res)) ##Adds LT estimates to the experimental setup data.frame setup <- cbind(setup,res[match(setup$well,rownames(res)),]) ## Time of injection is added to LT, so that LT really corresponds to time post injection time <- with(setup,strptime(as.character(time_injection),format="%H:%M:%S")) time <- as.numeric((max(time)-time)/(60^2)) setup$LT <- setup$LT+time ## Survival curves by dilution of bacterial culture injected library(survival) plot(survfit(Surv(LT,dead)~dilution,data=setup), lwd=c(3:1,1),lty=c(1,1,1,2), xlab="hours post injection",ylab="proportion of surviving insects") abline(h=0.5,col="red") legend("topright",lwd=c(1,3:1),lty=c(2,1,1,1),legend=c("LB",10^(3:1))) ## When does scepticemia start? res <- sapply(l,function(x) when.threshold(x$t_1_485_535,log(x$value_1_485_535,10), threshold=0.1)) setup$T_gfp <- res[match(setup$well,names(res))] setup$T_gfp <- setup$T_gfp + time ## Relation between time of death and moment when scepticemia is detected. ## Only points where scepticemia has been detected are represented here. plot(LT~T_gfp,data=setup,col=ifelse(dead,1,2),pch=as.numeric(dilution)) abline(0,1) with(setup,legend("topleft",legend=levels(dilution),pch=1:4)) # Most insects have died after scepticemia has started. }##Loading data data("galleria") ## dataset may contain NA if microplate reader has been stoped before the programmed ## end of the experiment d <- subset(galleria,!is.na(value)) l <- lapply(split(d,d$well),extract.well) data(setup) setup <- setup[match(setup$well,names(l)),] ## Computes rough estimates of variance in autofluorescence signal for dead and living insects ## These values serve as initial guess to fit the model. offset.alive <- estimate.offset(l,"1_330_405",min.t=1,max.t=5) # all insects are assumed to be alive during the # first five hours that follow injection offset.dead <- estimate.offset(l[which(setup$dead==1)],"1_330_405",min.t=72-5) # insects that were dead at the end of the experiment are # assumed to have died earlier than 5 hours before then # end of the experiment ## LT estimation or for a single well ## Check out well D9, to see what happens when an insect stayed alive. well <- "A3" plot(log(value_1_485_535,10)~t_1_485_535,type="l",col="green",ylim=c(2,5),data=l[[well]]) points(log(value_1_330_405,10)~t_1_330_405,type="l",col="gray",data=l[[well]]) # Rough estimate obtained using no offsets est1 <- with(l[[well]], estimate.LT(value_1_330_405,t_1_330_405,threshold.value=3)) abline(v=est1[["LT"]],lty=2,col="red") # Much better estimate obtained using offset for dead insects est2 <- with(l[[well]], estimate.LT(value_1_330_405,t_1_330_405, offset.dead = offset.dead,threshold.value=3)) abline(v=est2["LT"],lty=3,col="red") # Using offset.alive does not change anything to the estimate for well A3 # It may help for insect that have a larger variance in signal than others even after death est3 <- with(l[[well]], estimate.LT(value_1_330_405,t_1_330_405, offset.dead = offset.dead, offset.alive = offset.alive,threshold.value=3)) abline(v=est3["LT"],col="red") # Detection of significant GFP fluorescence (i.e. log fluorescence exceed # by 5% the maximum value observed during the first five hours) with(l[[well]],abline(v=when.threshold(t_1_485_535,log(value_1_485_535,10), threshold=0.1),col="green")) ##LT estimation for all wells if(FALSE) { #example takes time! Set to TRUE if you want to run it res <- sapply(l,function(x) estimate.LT(x$value_1_330_405,x$t_1_330_405, threshold.value=2,offset.dead=offset.dead, offset.alive=offset.alive)) res <- as.data.frame(t(res)) ##Adds LT estimates to the experimental setup data.frame setup <- cbind(setup,res[match(setup$well,rownames(res)),]) ## Time of injection is added to LT, so that LT really corresponds to time post injection time <- with(setup,strptime(as.character(time_injection),format="%H:%M:%S")) time <- as.numeric((max(time)-time)/(60^2)) setup$LT <- setup$LT+time ## Survival curves by dilution of bacterial culture injected library(survival) plot(survfit(Surv(LT,dead)~dilution,data=setup), lwd=c(3:1,1),lty=c(1,1,1,2), xlab="hours post injection",ylab="proportion of surviving insects") abline(h=0.5,col="red") legend("topright",lwd=c(1,3:1),lty=c(2,1,1,1),legend=c("LB",10^(3:1))) ## When does scepticemia start? res <- sapply(l,function(x) when.threshold(x$t_1_485_535,log(x$value_1_485_535,10), threshold=0.1)) setup$T_gfp <- res[match(setup$well,names(res))] setup$T_gfp <- setup$T_gfp + time ## Relation between time of death and moment when scepticemia is detected. ## Only points where scepticemia has been detected are represented here. plot(LT~T_gfp,data=setup,col=ifelse(dead,1,2),pch=as.numeric(dilution)) abline(0,1) with(setup,legend("topleft",legend=levels(dilution),pch=1:4)) # Most insects have died after scepticemia has started. }
Estimate variance in signal
estimate.offset(l, var, min.t = NULL, max.t = NULL)estimate.offset(l, var, min.t = NULL, max.t = NULL)
l |
A list object with each element being measurement for a sample |
var |
A character string corresponding to the name of the signal used to estimate lethal time |
min.t |
The time after which signal will be used to compute offset |
max.t |
The time before which signal will be used to compute offset |
A value that can be used as an offset in estimate.LT
Extract data for a single well
extract.well(data)extract.well(data)
data |
A |
The number of measurements might differ between wavelengths. In the Biotek reader used here, this can happen if fluorescence value exceeds the maximum value of 10^6. NA are then added to the data.frame.
A data.frame
The fluoSurv package provides functions to import fluorescence data as exported from a BioTek microplate reader and functions to estimate insect survival from these fluorescence data.
A dataset containing fluoresence measurements produced by a BioTek microplate reader. Measurements are taken from 96 larvae of Galleria mellonella which have been injected with a culture of the bacterium Xenorhabdus nematophila.
A data frame with 382944 rows and 8 variables:
Well name.
Intensity measurement.
Time.
Is the insect dead at the end of the experiment?
Read number, usually 1. This number will be greater than one when a combination of excitation and emission wavelengths is measured several times, with different gains.
Excitation wavelength.
Emission wavelength.
ID of the read. Combines read number and wavelengths to produce a unique ID for each set of fluorescence measurement. For example, if GFP fluoerscence has been measured with two different gains, the two measurements will be 1_485_535 and 2_485_535.
Reads a kinetic file, as produced by a Biotek plate reader.
read.kinetic(name, path = NULL, readTime = TRUE, saveData = TRUE)read.kinetic(name, path = NULL, readTime = TRUE, saveData = TRUE)
name |
The name of the file to be read |
path |
The path where the file is to be found |
readTime |
Should time data be read? |
saveData |
Should the resulting |
Returns a data.frame if saveData is set to FALSE.
If saveData is set to TRUE, the data.frame is saved and the
file name is returned.
## reads data. Warning: files are large, and this operation takes time! d <- read.kinetic("kinetics_xenorhabdus_galleria.txt", path=system.file('extdata', package = 'fluoSurv'), saveData=FALSE) str(d) ## saveData should rather be set to TRUE so that converted data are saved ## in a csv file and can be re-used later on.## reads data. Warning: files are large, and this operation takes time! d <- read.kinetic("kinetics_xenorhabdus_galleria.txt", path=system.file('extdata', package = 'fluoSurv'), saveData=FALSE) str(d) ## saveData should rather be set to TRUE so that converted data are saved ## in a csv file and can be re-used later on.
A dataset containing the description of an experimental setup where 96 larvae of Galleria mellonella have been injected with a culture of the bacterium Xenorhabdus nematophila.
A data frame with 96 rows and 4 variables:
Well name.
Dilution factor (log-transformed) of the injected culture. 1 therefore means 10 fold dilution, while LB corresponds to negative control where insects have been injected with sterile LB culture medium.
Time of the injection.
Is the insect dead at the end of the experiment?
Computes when fluorescence exceeds a given threshold value
when.threshold(t, x, min.t = 5, threshold = 0.1, n = 50)when.threshold(t, x, min.t = 5, threshold = 0.1, n = 50)
t |
The time value |
x |
The fluorescence value |
min.t |
The time value after which threshold value is searched |
threshold |
Threshold value, as a proportion above the maximum intensity value observed before min.t |
n |
Width of the moving average window used to smooth signal |
data(galleria) d <- subset(galleria,!is.na(value)) l <- lapply(split(d,d$well),extract.well) #complete kinetics for each well with(l[["A3"]],plot(t_2_485_535,log(value_2_485_535,10),type="l")) with(l[["A3"]],abline(v=when.threshold(t_2_485_535,value_2_485_535)))data(galleria) d <- subset(galleria,!is.na(value)) l <- lapply(split(d,d$well),extract.well) #complete kinetics for each well with(l[["A3"]],plot(t_2_485_535,log(value_2_485_535,10),type="l")) with(l[["A3"]],abline(v=when.threshold(t_2_485_535,value_2_485_535)))