respR (version 2.3.3)

auto_rate.int: Run auto_rate on multiple replicates in intermittent-flow respirometry data

Description

auto_rate.int allows you to run the auto_rate() function on multiple replicates in intermittent-flow respirometry. A wait and measure phase can be specified for each replicate, and the auto_rate analysis is performed within the measure region.

Usage

auto_rate.int(
  x,
  starts = NULL,
  wait = NULL,
  measure = NULL,
  by = "row",
  method = "linear",
  width = NULL,
  n = 1,
  plot = TRUE,
  ...
)

Value

Output is a list object of class auto_rate.int containing a auto_rate object for each replicate in $results. The output also contains a $summary table which includes the full rate regression results from each replicate with replicate number indicated by the $rep column. Output also contains a $rate element which contains the rate values from each replicate in order. The function call, inputs, and other metadata are also included. Note, that if you have many replicates this object can be rather large (several MB).

Arguments

x

object of class inspect or data.frame. This is the timeseries of paired values of oxygen against time containing multiple replicates from which to calculate rates.

starts

Numeric. Row locations or times (in the units of the data in x) of the start of each replicate. If a single value it indicates a regular interval in rows or time starting from row 1. If a vector, each entry is the start row or time of an individual replicate. Use of rows or time is controlled via by.

wait

Numeric. A row length or time duration to be applied at the start of each replicate to exclude these data from any rate calculations. Can be a single value to apply the same wait phase to each replicate, or a vector of the same length as starts of different wait phases for each replicate. Optional.

measure

Numeric. A row length or time duration to be applied at the end of the wait phase (if used), and used to exclude the flush period. This is the region within which the auto_rate analysis is conducted for each replicate. Can be a single value to apply the same measure phase to each replicate, or a vector of the same length as starts of different measure phases for each replicate. Default is NULL in which case the entire replicate is used (which is rarely what is wanted).

by

String. "row" or "time". Controls how starts, wait and measure are applied. It also controls how the width is applied in the auto_rate analysis - see help("auto_rate"). Default is "row".

method

string. The auto_rate method to use. Default is "linear". Others include "lowest" and "highest". See help("auto_rate") for descriptions and other methods.

width

numeric. The width to use in the auto_rate analysis. Mandatory and should be entered in the correct units of the by input. See help("auto_rate") and vignettes on website for how width affects analyses.

n

integer. How many auto_rate results to return for each replicate. Default is 1.

plot

logical. Default is TRUE. Plots the results. See 'Plotting' section for details.

...

Allows additional plotting controls to be passed, such as type, pos, legend, and quiet.

Details

auto_rate.int uses the starts input to subset each replicate. The wait and measure inputs control which parts of each replicate data are excluded and included from the rate calculation. It runs auto_rate on the measure phase in each replicate saving the top n ranked results and extracting the rate and other data to a summary table.

The x input should be aninspect object. Alternatively, it can be a two-column data frame containing paired values of time and oxygen from an intermittent-flow experiment in columns 1 and 2 respectively (though we always recommend processing such data in inspect() first). If a multiple column dataset is entered as x the first two columns are selected by default. If these are not the intended data use inspect to select the correct time and oxygen columns.

auto_rate inputs

You should be familiar with how auto_rate works before using this function. See help("auto_rate") and vignettes on the website for full details.

The auto_rate inputs can be changed by entering different method and width inputs. The by input controls how the width is applied. Note if using a proportional width input (i.e. between 0 and 1 representing a proportion of the data length) this applies to the length of the measure phase of each particular replicate.

The n input controls how many auto_rate results from each replicate to return in the output. By default this is only the top ranked result for the particular method, i.e. n = 1. This can be changed to return more, however consider carefully if this is necessary as the output will necessarily contain many more rate results which may make it difficult to explore and select results (although see select_rate()).

Specifying replicate structure

The starts input specifies the locations of the start of each replicate in the data in x. This can be in one of two ways:

• A single numeric value specifying the number of rows in each replicate starting from the data in the first row. This option should only be used when replicates cycle at regular intervals. This can be a regular row or time interval, as specified via the by input. If the first replicate does not start at row 1, the data should be subset so that it does (see subset_data()) and example here. For example, starts = 600, by = "row" means the first replicate starts at row 1 and ends at row 600, the second starts at row 601 ends at 1200, and so on.

• A numeric vector of row locations or times, as specified via the by input, of the start of each individual replicate. The first replicate does not have to start at the first row of the data, and all data after the last entry is assumed to be part of the final replicate. Regular R syntax such as seq(), 1:10, etc. is also accepted, so can be used to specify both regular and irregular replicate spacing.

For both methods it is assumed each replicate ends at the row preceding the start of the next replicate, or in the case of the last replicate the final row of the dataset. Also for both methods, by = "time" inputs do not need to be exact; the closest matching values in the time data are used.

Results are presented in the summary table with rep and rank columns to distinguish those from different replicates and their ranking within replicates (if multiple results per replicate have been returned by increasing the n input).

Specifying rate region

The wait and measure inputs are used to specify the region from which to extract a rate and exclude flush periods. They can be entered as row intervals or time values in the units of the input data. The wait phase controls the amount of data at the start of each replicate to be ignored, that is excluded from any rate calculations. The measure phase determines the region after this from which a rate is calculated. Unlike calc_rate.int(), auto_rate.int will not necessarily use all of the data in the measure phase, but will run the auto_rate analysis within it using the method, width and by inputs. This may result in rates of various widths depending on the inputs. See auto_rate() for defaults and full details of how selection inputs are applied.

There is no flush phase input since this is assumed to be from the end of the measure phase to the end of the replicate.

Both wait and measure can be entered in one of two ways:

• Single numeric values specifying a row width or a time period, as specified via the by input. Use this if you want to use the same wait and measure phases in every replicate.

• If starts is a vector of locations of the start of each replicate, these inputs can also be vectors of equal length of row lengths or time periods as specified via the by input. This is only useful if you want to use different wait and/or measure phases in different replicates.

If wait = NULL no wait phase is applied. If measure = NULL the data used for analysis is from the start of the replicate or end of the wait phase to the last row of the replicate. This will typically include the flush period, so is rarely what you would want.

Example

See examples below for actual code, but here is a simple example. An experiment comprises replicates which cycle at ten minute intervals with data recorded every second. Therefore each replicate will be 600 rows long. Flushes of the respirometer take 3 minutes at the end of each replicate. We want to exclude the first 2 minutes (120 rows) of data in each, and run an auto_rate analysis to get an oxygen uptake rate within the following five minute period (300 rows), leaving the three minutes of flushing (180 rows) excluded. The inputs for this would be:

starts = 600, wait = 120, measure = 300, by = "row"

Plot

If plot = TRUE (the default), the result for each rate is plotted on a grid up to a maximum of 20. There are three ways of plotting the results, which can be selected using the type input:

• type = "rep": The default. Each individual replicate is plotted with the rate region highlighted in yellow. The wait and measure phases are also highlighted as shaded red and green regions respectively. These are also labelled if legend = TRUE.

• type = "full": Each replicate rate is highlighted in the context of the whole dataset. May be quite difficult to interpret if dataset is large.

• type = "ar": Plots individual replicate results as auto_rate objects. Note, these will only show the measure phase of the data.

For all plot types pos can be used to select which rate(s) to plot (default is 1:20), where pos indicates rows of the $summary table (and hence which $rep and $rank). This can be passed either in the main function call or when calling plot() on output objects. Note for all plot types if n has been changed to return more than one rate per replicate these will also be plotted.

S3 Generic Functions

Saved output objects can be used in the generic S3 functions plot(), print(), summary(), and mean(). For all of these pos selects rows of the $summary table.

• plot(): plots the result. See Plot section above.

• print(): prints the result of a single rate, by default the first. Others can be printed by passing the pos input. e.g. print(x, pos = 2)

• summary(): prints summary table of all results and metadata, or the rows specified by the pos input. e.g. summary(x, pos = 1:5). The $rep column indicates the replicate number, and $rank column the ranking of each rate within each replicate (only used if a different n has been passed, otherwise they are all 1). The summary table (or pos rows) can be exported as a separate data frame by passing export = TRUE.

• mean(): calculates the mean of the rates from every row or those specified by the pos input. e.g. mean(x, pos = 1:5) Note if a different n has been passed this may include multiple rates from each replicate. The mean can be exported as a numeric value by passing export = TRUE.

More

For additional help, documentation, vignettes, and more visit the respR website at https://januarharianto.github.io/respR/

Examples

# \donttest{
# Irregular replicate structure ------------------------------------------

# Prepare the data to use in examples
# Note in this dataset each replicate is a different length!
data <- intermittent.rd
# Convert time to minutes (to show different options below)
data[[1]] <- round(data[[1]]/60, 2)
# Inspect
urch_insp <- inspect(data)

# Calculate the most linear rate within each replicate
auto_rate.int(urch_insp,
              starts = c(1, 2101, 3901),
              by = "row",
              method = "linear",
              width = 400) %>%
  summary()

# Calculate the lowest rate within each replicate across
# 5 minutes (300 rows). For this we need to specify a 'measure' phase
# so that the flush is excluded.
auto_rate.int(urch_insp,
              starts = c(1, 2101, 3901),
              measure = 1000,
              by = "row",
              method = "lowest",
              width = 300) %>%
  summary()

# You can even specify different 'measure' phases in each rep
auto_rate.int(urch_insp,
              starts = c(1, 2101, 3901),
              measure = c(1000, 800, 600),
              by = "row",
              method = "lowest",
              width = 300) %>%
  summary()

# We usually don't want to use the start of a replicate just after the flush,
# so we can specify a 'wait' phase. We can also specify 'starts', 'wait',
# 'measure', and 'width' in units of time instead of rows.
#
# By time
# (this time we save the result)
urch_res <- auto_rate.int(urch_insp,
                          starts = c(0, 35, 65), # start locations in minutes
                          wait = 2,              # wait for 2 mins
                          measure = 10,          # measure phase of 10 mins
                          by = "time",           # apply inputs by time values
                          method = "lowest",     # get the 'lowest' rate...
                          width = 5) %>%          #  ... of 5 minutes width
  summary()

# Regular replicate structure --------------------------------------------

# If replicates cycle at regular intervals, 'starts' can be used to specify
# the spacing in rows or time, starting at row 1. Therefore data must be
# subset first so that the first replicate starts at row 1.
#
# Subset and inspect data
zeb_insp <- zeb_intermittent.rd %>%
  subset_data(from = 5840,
              to = 75139,
              by = "row",
              quiet = TRUE) %>%
  inspect()

# Calculate the most linear rate from the same 6-minute region in every
# replicate. Replicates cycle at every 660 rows.
zeb_res <- auto_rate.int(zeb_insp,
                         starts = 660,
                         wait = 120, # exclude first 2 mins
                         measure = 360, # measure period of 6 mins after 'wait'
                         method = "linear",
                         width = 200, # starting value for linear analysis
                         plot = TRUE) %>%
  summary()

# S3 functions ------------------------------------------------------------

# Outputs can be used in print(), summary(), and mean().
# 'pos' can be used to select replicate ranges
summary(zeb_res)
mean(zeb_res, pos = 1:5)

# There are three ways by which the results can be plotted.
# 'pos' can be used to select replicates to be plotted.
#
# type = "rep" - the default. Each replicate plotted on a grid with rate
# region highlighted (up to a maximum of 20).
plot(urch_res)

# type = "full" - each replicate rate region plotted on entire data series.
plot(urch_res, pos = 1:2, type = "full")
# Of limited utility when datset is large
plot(zeb_res, pos = 10, type = "full")

# type = "ar" - the 'auto_rate' object for selected replicates in 'pos' is plotted
# Note this shows the 'measure' phase only
plot(urch_res, pos = 2, type = "ar")

# See vignettes on website for how to adjust and convert rates from auto_rate.int
# }