Changelog

Source: NEWS.md

chromatographR 0.9.0

Breaking changes

Preprocessing

• preprocess no longer clamps negative values to zero by default. Set the new zero_floor = TRUE for bit-identical previous output.
• subtract_blanks no longer clamps negative values to zero by default. Set zero_floor = TRUE for the previous output; it now warns when it zeroes more than 5% of a sample.
• subtract_blanks now matches blanks to samples by retention time, interpolating across differing grids. Previously a blank of another length errored, and one of equal length was subtracted row by row, even when its times were offset.
• Deprecated preprocess’s remove.time.baseline argument in favor of baseline = "none".
• maxI now follows ... in preprocess’s signature, so it must be named in full. Partial matching otherwise captured m, which belongs to the baseline estimators. maxI can no longer be given positionally.
• preprocess now validates dim1 and dim2, which must be strictly increasing. A named by, length.out, or along.with element gets an informative error pointing at seq().

Retention time correction

• Inverted the meaning of correct_rt’s penalty argument for alg = "vpdtw", so a higher value now allows less warping. The default changed from 5 to 0.2; to convert an explicit value, take its reciprocal. Values above 1 now warn.
• The trwdth argument in correct_rt now specifies a width in minutes, not a count of timepoints. Multiply old values by the sampling interval in minutes to convert it; a value wider than the chromatogram now errors.
• correct_rt now passes trwdth to ptw, which previously ignored it. The default of 0.2 matches ptw’s own 20 timepoints at dt = 0.01.
• max_shift defaults to 1 and dilation_span to 0.3, so VPdtw alignments change. See ?correct_rt.
• correct_rt(scale = TRUE) now scales chromatograms only to fit the warp, so corrected values keep their original units. Scaling divides by the maximum instead of rescaling to [0, 1]. Use preprocess(maxI = 1) to rescale the data.
• correct_rt and correct_rt_group now error when samples’ time axes differ by a sampling interval or more. Previously they warped such samples by timepoint index. Interpolate onto a common grid with preprocess(dim1 = ...) first.
• correct_rt and correct_rt_group now error on duplicate sample names, as get_peaks already did.

Peak detection and integration

• fit = "bemg" now reports tau_left and tau_right as tail time constants in time_unit, so a larger value means a longer tail. In 0.8.0 they were rates (1/time), with left and right swapped.
• Corrected an off-by-one error in find_peaks that reported every apex one timepoint early. Retention times from get_peaks and get_peaktable shift accordingly, and peaks whose preceding timepoint had been clipped to zero by preprocess are now detected.
• Fitted retention times can now fall between timepoints, so rt can shift by up to one timepoint and "bemg" no longer understates height. correct_peaks no longer rounds rt.cor for ptw or lptw models. Match rt to the time axis with a tolerance, not with == or match().
• With baseline = "sloped", the fitted local baseline no longer rises above the signal at the edges of the peak window. Previously it could, underestimating "egh" and "gaussian" areas.
• r.squared in get_peaks and fit_peaks output is now 1 - SS_res / SS_tot, not the squared correlation between the fit and the data. It is lower for poor fits and can be negative, so revisit any cutoffs.
• plot_spectrum and plot_all_spectra now label the correct spectral maxima, having inherited the same off-by-one error. Reported lambda-max values shift by one grid step.
• find_peaks now rounds the "mva" and "tmva" smoothing windows up to an odd width of at least three points, which changes which apices are detected for these methods.
• find_peaks now errors when smooth_window is narrower than seven timepoints under Savitzky-Golay ("savgol") smoothing. Previously it returned the signal unsmoothed.
• find_peaks and get_peaks now smooth over 3 timepoints by default. The old default, smooth_window = 0.001, was a proportion of the run, so the smoothing varied with run length. Set it explicitly to restore the old behavior; see Choosing the smoothing window in ?find_peaks.
• Renamed find_peaks’s slope_thresh and amp_thresh arguments to min_curvature and min_amplitude. slope_thresh was a misnomer because the argument thresholded the second derivative, not the slope. For now, the old names still work with a deprecation warning.
• get_peaks and fit_peaks no longer filter on peak width by default. Previously sd_max dropped peaks with sd > 50 timepoints; to filter by width, call filter_peaks(max_fwhm = ) on the result.

Peak purity

• Overhauled get_purity, correcting the purity ratio, so purity values change for every peak. cutoff now applies to the background-corrected trace, as a proportion of peak height rather than of total absorbance. See ?get_purity.
• Added arguments to get_purity: background_correction, reference, metric, min_run and max_threshold control the purity calculation, and noise_method, noise_quantile and noise_window control noise estimation.
• Deprecated the weight argument to get_purity and get_agilent_threshold in favor of noise_inflation.
• Removed the noise_threshold argument from get_peaks and fit_peaks, which had no effect. Pass purity options as purity_args instead.
• fit_peaks now defaults to estimate_purity = FALSE, matching get_peaks, so its purity column is NA unless you set estimate_purity = TRUE.
• plot_purity = TRUE in plot.peak_list now draws dissimilarity and threshold curves. It also accepts "curves_local" and "ratio"; see ?plot.peak_list.

Peak tables and plotting

• Retention times in dimnames are now formatted to a consistent number of decimal places, so an interpolated time axis is no longer ragged. Lookups by the old names need updating (x["10.10", ] rather than x["10.1", ]).
• plot.peak_list’s y.offset is now a fraction of the plot height, with a default of 0.03. Previously it was in absorbance units, with a default of 25. Values above 1 warn. Boundary ticks are scaled the same way.
• get_peaktable’s hmax and sigma_t, and combine_peaks’s tol, are now in minutes, regardless of the time unit of the data. Results change only for chromatograms whose times are not in minutes.
• reshape_chroms now reads time_resolution in minutes, matching plot_chroms. Previously it was in the units of the time axis, so divide old values by 60 for chromatograms in seconds.
• Removed get_peaktable’s out argument, so peak tables are always data.frames. Use as.matrix(pt$tab) for a matrix.
• Peak tables built from corrected peaks now name the pk_meta rows rt, start and end rather than rt.cor, start.cor and end.cor, and so does reshape_peaktable. Check args$use_corrected_rt to see which times were used.
• attach_metadata now adds to an existing sample_meta, replacing only the columns of the same name. Previously it replaced the whole table. To do that now, set peak_table$sample_meta <- NA first.

Example data

• Regenerated the example datasets: Sa_pr and Sa_warp are no longer clipped at zero, and most pk_tab peak names have shifted; for example, V13 is now V11.

Other renamed and deprecated arguments

• Renamed dot-separated argument names to snake_case: spec.smooth to smooth_spectra, min.cor to min_cor, n.traces/n.zeros to n_traces/n_zeros, sigma.t/sigma.r to sigma_t/sigma_r, and use.cor to use_corrected_rt. The old names still work, with a warning. init.coef and max.only keep their names, since they are passed through to external functions that define them.
• Deprecated sd_max in get_peaks and fit_peaks in favor of filter_peaks(max_fwhm = ), and min_sd/max_sd in filter_peaks in favor of min_fwhm and max_fwhm. A standard deviation is equivalent to peak width only for fit = "gaussian". See ?filter_peaks.
• Renamed the Whittaker smoothing penalty passed to preprocess and correct_baseline from lambda to stiffness, reserving lambda for the common usage to specify a wavelength used throughout the package.
• Removed the lambda argument from get_spectra, which had no effect.
• Renamed correct_rt’s maxshift to max_shift, now in minutes. maxshift still works, read as a count of timepoints as before, with a warning.

New features

New functions

• Added correct_baseline function for estimating and subtracting a baseline from a chrom_list along the time axis. Unlike preprocess it does not interpolate, smooth or rescale, so it can be applied to data that are already preprocessed or aligned.
• Added get_baseline and plot_baseline for retrieving and plotting the baseline retained by preprocess or correct_baseline when attach_baseline = TRUE.
• Added crop_chroms function for restricting chromatograms to a range of retention times (rt_range) and/or wavelengths (lambda_range). Cropping is pure subsetting, so values and dimnames pass through unchanged, unlike preprocess(dim1 = ), which interpolates onto a new axis.
• Added match_peaks function for matching peaks across chromatograms using spectral correlation and retention time proximity. The resulting matched_peak_list can be passed to correct_rt for landmark-based PTW or to get_peaktable. See ?targeted_analysis for matching to a table of targets.
• Added get_time_resolution for reading a chromatogram’s sampling interval, in the units of its time axis by default or in minutes via unit.
• Added plot_chrom_contour for plotting one chromatogram across retention time and wavelength, on a compressed color scale by default.

New generic functions

• Converted get_peaktable to an S3 generic. The default method keeps the original behavior, including for unclassed lists. The new get_peaktable.matched_peak_list method builds a peak table directly from matched peak pairs.
• Added a plot method for chrom_list objects from read_chroms, drawing traces, a heatmap or a contour plot according to type.

Baseline correction

• Added a baseline argument to preprocess choosing the time-axis baseline estimator: "als" (the default), "arpls", "aspls" or "snip". See ?baseline_estimators.
• The smoothing parameter each estimator used is recorded on the result as a "baseline_param" attribute, carrying the estimator’s name.
• Added an attach_baseline argument to preprocess, keeping the estimated baseline as an attribute for inspection. It is removed by correct_rt and subtract_blanks, whose output it would no longer match.
• Added reference-wavelength subtraction to preprocess: reference_wavelengths removes an offset shared by every wavelength before the baseline is estimated.
• Baseline estimators now require a complete trace, and give an informative error when the signal contains missing or non-finite values. Previously a gap either caused an obscure error or had its two sides spliced together.

Retention time correction

• Added landmark parametric time warping (alg = "lptw") to correct_rt and correct_rt_group, which fits the warp to matched peak positions rather than whole signals. See ?correct_rt.
• Added a dilation_span argument to correct_rt: the width, in minutes, of the moving-maximum window used to build the "vpdtw" penalty, previously hardcoded to 350 timepoints. A span wide enough to make the penalty constant now warns.
• correct_rt_group can now return its warping models, one per sample, with what = "models", for use with correct_peaks.
• correct_peaks now accepts "lptw" and "vpdtw" models from correct_rt, not only "ptw".
• Added a chrom_list_aligned argument to correct_peaks, naming aligned chromatograms for later spectrum and trace lookups.

Peak detection and integration

• Added a min_prominence_snr argument to find_peaks, filtering apices by topographic prominence in multiples of the estimated noise. Unlike min_curvature and min_amplitude this threshold is scale-free, so it transfers between datasets. find_peaks now also returns a prominence column.
• Added manual integration of fixed retention-time windows to get_peaks via a windows argument, quantifying a compound even where no peak is detected. get_peaktable gives one column per window. See ?get_peaks.
• Added targets, tol and select_by arguments to get_peaks for fitting only the peaks near target retention times, which is faster than fitting every peak. get_peaktable groups the result by target. See ?targeted_analysis.

Fit diagnostics

• Two fit-diagnostic columns are now reported by get_peaks: converged (flagging peaks that didn’t converge) and window_limited (flagging peaks whose fitted shape sits at a limit set by the integration window or the sampling interval).
• Fit diagnostics can now be used for filtering: filter_peaks gains converged and window_limited arguments, and filter_peaktable gains min_converged. All three error on fit = "raw", which reports no fit diagnostics.
• get_peaks now warns when more than a fifth of the peak list is integration windows too small to fit, which usually means find_peaks is detecting noise. Suppressed for fit = "raw" and when windows is supplied.

Peak purity

• Purity estimation in get_peaks is now much faster, so it is no longer a bottleneck.
• Purity estimation now warns when a chromatogram appears to be clipped at zero, which makes peaks look less pure. Use zero_floor = FALSE, the default in both preprocess and subtract_blanks.

Peak tables and plotting

• Added targets, tol and pad arguments to plot.peak_list for zooming in on the peaks near given retention times, rescaling both axes to the selected window.
• plot.peak_list now draws axes and titles each plot with its sample, wavelength and peak model. Pass xaxt = "n", yaxt = "n" or main = "" to remove them.
• plot.peak_list now accepts numbers = "rt", labeling peaks with their retention times, which can be passed to targets to zoom in on them.
• Added a what argument to plot_chroms for plotting one element of nested chromatograms from read_chroms, such as "dad" or a pump pressure trace.
• Added a zlim argument to plot_chroms_heatmap, fixing the range of its color scale so that several plots can share one scale. By default the scale now spans the plotted window in every engine; previously the "ggplot" and "plotly" engines used the whole run when xlim was set.
• Added format = "wide" to reshape_peaktable, returning one row per sample with metadata and one column per selected peak.
• annotate_peaks now accepts retention times through the loc argument, labeling the highest point within tol of each supplied retention time without requiring a peak table. See ?annotate_peaks.
• get_peaktable now fills sample_meta with any fields added to the chromatograms with chromConverter::add_metadata (chromConverter 0.11.0 or later).
• filter_peaks and filter_peaktable now record each call as code, which print shows for peak lists and get_peaktable stores in args$filters. See ?filter_peaks.
• get_peaktable now records how the peaks were fitted (fit, baseline and max_iter) in args, and keeps any windows or targets as attributes, including targets found in no sample. See ?get_peaktable.
• cluster_spectra now numbers and colors each box on the dendrogram to match its spectra panel, with colors set by col, and boxes only clusters passing min_size and max_size.

Bug fixes and other minor changes

Preprocessing and baseline correction

• preprocess no longer fails on chromatograms with a few discontinuous wavelengths — smooth.spline requires at least four unique values, so spectral smoothing is now skipped below four wavelengths.
• preprocess now rejects a single-valued dim1 or dim2 when interpolating onto it, instead of failing later with an opaque dimnames error.
• preprocess now reads the time unit after unwrapping nested input, so outlier_cutoff is scaled correctly for chromatograms supplied as list(dad = <matrix>).
• preprocess now ignores missing values when rescaling to maxI.
• Spectral smoothing in preprocess now accounts for the actual spacing of the wavelength axis rather than treating the columns as evenly spaced. Results are unchanged for evenly spaced grids, including all of the bundled data.
• Corrected the documented units of outlier_cutoff in preprocess, which is expressed in minutes.
• Corrected the documented order of operations in preprocess: the code interpolates first, then smooths spectrally, then baselines.
• preprocess, correct_rt and get_peaks now stop with an error when a sample fails in parallel (cl > 1). Previously they returned try-error or NULL elements, including for samples that had succeeded but were processed by the failing worker.
• Warnings raised while processing samples in parallel (cl > 1) now reach the user. Each distinct warning is raised once, naming the samples that raised it.

Retention time correction

• correct_rt(alg = "vpdtw") now reproduces the alignment VPdtw fitted. Previously samples could remain misaligned when the warp extended past the end of the reference.
• correct_rt no longer misplaces peaks when n_zeros > 0.
• correct_peaks no longer errors on models fitted by correct_rt with scale = TRUE or n_zeros > 0 when chrom_list is omitted, and now gives a clear error for samples with no warping model and for peak lists built with collapse = TRUE.
• correct_rt(models = ...) now matches models to samples by name. Previously models were applied by position, so reordering chrom_list warped each sample with another sample’s model.
• correct_peaks now replaces the corrected times of an already corrected peak list. Previously it added a second set, and get_peaktable kept using the first.
• correct_rt and correct_rt_group now return a chrom_list when given one, so plot() and chromConverter’s print method still apply after alignment.

Peak detection and integration

• get_peaks no longer inflates the areas of peaks bordering missing values, such as the padded ends of chromatograms aligned by correct_rt(alg = "vpdtw"), which could be several times too large.
• get_peaks now converts sd, FWHM, tau and area correctly when the time axis is not in minutes. time_unit now defaults to the chromatogram’s own unit, so default output is unchanged.
• find_peaks(bounds = TRUE) now returns its full four-column schema when no peaks are detected, rather than dropping lower and upper.
• max_iter now controls how long the peak fitters run. Previously it was ignored, leaving nls.lm’s default of 100 * (parameters + 1) evaluations. "egh" and "bemg" fits may shift slightly.
• get_peaks no longer errors when a sample contains no peaks; previously a single empty sample failed the whole call. fit_peaks now returns an empty data.frame rather than a one-row matrix of NAs.
• Printing a peak_list no longer errors when a sample has no peaks at any wavelength, as in peak lists read by chromConverter::read_peaklist.
• FWHM is now computed as 2*sqrt(2*log(2))*sigma rather than the rounded 2.35*sigma.
• Requesting a wavelength missing from a single-column chromatogram now errors, instead of returning the existing column falsely labeled with the requested wavelength.
• fit = "bemg" no longer returns its starting values in place of a fit when a peak’s integration window spans about 700 timepoints or more.
• fit = "bemg" now rejects fits whose tail grows away from the apex, which inflated the reported area. Such peaks are refitted, and flagged converged = FALSE if the refit also fails.
• fit = "gaussian" and "egh" peaks can no longer narrow below one timepoint at half height. Previously a fit to a noise apex could collapse between two samples and report a height several times the signal around it.

Peak tables and filtering

• get_peaktable no longer spreads pk_meta across several columns when all peaks at a wavelength fall in one cluster.
• get_peaktable no longer rounds pk_meta to two decimal places, which moved retention times by up to 0.005 minutes and set small tau values to zero.
• Peak tables built from correct_peaks output now read spectra and traces from aligned chromatograms, rebuilt from the stored warping models. Previously they were read from the unaligned data at aligned times. See ?correct_peaks.
• filter_peaktable no longer appends a spurious rt row to pk_meta when filtering a table built from corrected peaks.
• normalize_data no longer reports column as ambiguous when another column’s name merely contains it, such as "V12" for "V1".
• combine_peaks now removes dropped peaks from the ref_spectra element so that it stays aligned with tab.
• combine_peaks(choose = "least_sparse") now keeps the peak detected in the most samples. Previously every candidate tied and the first was kept.
• filter_peaktable now errors when rts is combined with min_rt or max_rt.
• write_peaktable(what = "args") now writes a CSV that reads back. Previously arguments holding several values, such as two or more wavelengths, spilled into extra columns.
• reshape_peaktable now keeps lambda and rt when peaks are renamed with a named peaks vector. Previously both were NA.
• reshape_peaktable now errors on metadata columns it cannot find, which were previously dropped, and gives clear errors for unattached metadata or unknown peaks.

Plotting

• annotate_peaks and plot_spectrum_inset now place labels at each peak’s maximum for tables built from corrected peaks, as for uncorrected ones.
• plot_all_spectra axes, and the absorbance axis of plot_spectrum’s ggplot2 and plotly traces, now show wavelength and absorbance units instead of the time unit.
• plot_all_spectra now names the returned spectra when idx gives sample names, and errors on names not in chrom_list.
• plot_all_spectra(what = "rt") and what = "idx" no longer require a peak_table, and a loc outside every chromatogram now gives a clear error.
• plot.peak_table(idx = "all") no longer errors when chrom_list is omitted, and attach_ref_spectra, mirror_plot and get_peaktable(clust = "sp") now treat chrom_list = NULL as omitted. Omitting lambdas for multi-wavelength data in mirror_plot now gives a clear error.
• plot_chroms(engine = "base") now extends the y axis below zero when the traces do. Previously it always started at 0, cutting off negative values.
• The base-engine legend in plot_chroms no longer covers the traces at a top or bottom legend_position, including the default "topright"; unless ylim is set, the y axis extends to make room.
• plot_chroms now fits the y axis to the traces within xlim when ylim is not given, under the "base" and "plotly" engines. Previously it spanned the whole run.
• plot_spectrum_inset now draws spectra from the supplied chrom_list when idx covers several samples. Previously it read the peak table’s own chromatograms instead.
• plot.peak_list(numbers = TRUE) now draws each number just above its peak. Previously the numbers were drawn at arbitrary heights.
• plot.peak_list now draws shaded peaks at the right width for peak lists built with get_peaks(time_unit = "s") or "ms". Previously they were drawn too wide.
• plot.peak_list now accepts xlab, ylab and type. Previously passing any of them gave an error.
• plot.peak_list no longer fails with “subscript out of bounds” when the wavelength given to get_peaks is written differently from the chromatogram’s column name, such as "220" for "220.00000".
• plot_chroms no longer errors on single-wavelength chromatograms under the "ggplot" and "plotly" engines. A wavelength given in lambdas for such data must now match the column name. Previously it was ignored.
• plot_chroms and reshape_chroms no longer fail with invalid '(to - from)/by' when the sampling interval is more than twice the requested time_resolution.
• reshape_chroms no longer errors when lambdas is left at its NULL default, which now returns every wavelength, or when sample_var is not "sample".
• plot_chroms now draws each wavelength as a separate line under the "ggplot" and "plotly" engines, instead of joining them into one.
• The plot.ptw_list heatmap now plots the wavelength in lambdas and puts both panels on one color scale. Previously it always used the first wavelength, the base engine showed only the warped panel, and the other engines’ legend matched only the raw panel.
• scan_chrom(export_spectrum = TRUE) now returns the spectrum instead of the scan index.
• plot.peak_table now warns when several loc are given but only the first is returned.
• plot_chroms_heatmap(engine = "base") now labels its color bar with the range of the plotted colors. Previously the labels came from the wrong rows and were rounded to the nearest 100, so they could all read 0.

Documentation and packaging

• Updated the GC-FID vignette to fetch its example data from a mirror on GitHub, since Dryad now requires authentication for all downloads.

chromatographR 0.8.0

Breaking changes

• Added support for bidirectional exponentially modified gaussian (BEMG) peak shape (fit = "bemg" in get_peaks/fit_peaks). The default peak shape remains "egh" for now but this may change in a future release.
• Added summarize_by argument to get_peaktable controlling how peak metadata are summarized within each cluster. The new default, "weighted.mean", weights peaks by intensity; "mean", "median" and "max" are also available. Set summarize_by = "mean" to reproduce previous behavior.
• Added automatic thinning of chromatograms in plot_chroms to speed up plotting of raw data, adjustable with the new time_resolution argument.
• Changed behavior of plot_chroms so the legend is no longer displayed by default.
• When export_spectrum == TRUE, plot_spectrum and plot_all_spectra now return spectra invisibly.
• Corrected get_peaks to return model-fitted peak heights rather than raw peak heights. This changes numerical output for workflows relying on the previous behavior.
• Added fill_zeros to correct_rt, controlling whether out-of-bounds regions produced by warping are filled with zeros or returned as NA. Defaults to FALSE, which changes VPdtw behavior.
• Changed the name of the a parameter in plot.peak_list to alpha.
• Renamed the chrom_list argument to x in plot_chroms_heatmap, which now also accepts a peak table in the first position.
• Removed index and chr arguments across plotting functions, previously deprecated in v0.7.0. Use idx instead.
• Corrected merge_peaks so the result no longer depends on the order in which peaks are supplied.
• get_purity now decides which timepoints belong to a peak from the wavelength of maximum absorbance at the apex, rather than the first wavelength in the chromatogram.
• Deprecated arguments in get_peaks still using dot notation (e.g., sd.max). Use the snake_case equivalents instead.
• Deprecated peak argument in plot_all_spectra in favor of loc.
• Deprecated the save argument in cluster_spectra in favor of outfile.
• Removed deprecated peak_no argument from cluster_spectra.

New features

New functions

• Added subtract_blanks function for subtracting blank chromatograms from a chrom_list.
• Added correct_rt_group function for group-wise warping function estimation using within-group averages. This improves alignment when samples share batch-level shifts or contain group-specific peaks.
• Added plot_spectrum_inset function for plotting UV spectra overlaid onto a chromatogram.
• Added annotate_peaks function for adding text labels to peaks in plots created by plot_chroms.
• Added get_spectra function for extracting spectra from chromatograms. The export_spectrum argument in plot_spectrum and plot_all_spectra is deprecated in favor of get_spectra.

New generic functions

• Added summary method for peak_table objects, reporting dimensions, retention time range, intensity statistics, and metadata availability.
• Added print.peak_table function to return just the peak data.
• Added [ method for peak_table objects for simultaneous subsetting of tab, pk_meta, sample_meta, and ref_spectra elements.
• Added [ method for peak_list objects to preserve metadata attributes when subsetting.
• Added print method for peak_list objects, summarizing samples, wavelengths, fit method, and peak counts.

Enhancements

• Added support for fitting a local baseline beneath each peak through the baseline argument in get_peaks: "none" (default), "flat" or "sloped". Peak heights and areas are computed above the fitted baseline.
• Added support for plotting spectra from multiple chromatograms in plot_spectrum by supplying a vector to idx (e.g. idx = c(1, 2, 4)). Overlaid traces and spectra are color-coded by sample.
• Added new options to combine_peaks via the choose argument: "least_sparse" retains the peak detected in the most samples, and "lambda" retains the peak measured at a preferred wavelength.
• Preprocess can now take nested lists containing a dad element in each list. This change allows better compatibility with chromConverter.
• Added more informative warnings and errors to preprocess function.
• Added support for numeric input to lambda argument in reshape_chroms.
• Refactored normalize_data so it detects metadata and peak table columns automatically; added an on_invalid argument controlling how zero, negative and NA normalization values are handled; and improved performance.
• Refactored ptw warping in correct_rt to interpolate the whole matrix at once instead of column by column.
• Added max_zeros argument in filter_peaktable to exclude features above a threshold proportion of zero values.
• Added support for new chrom_list class introduced in chromConverter v0.9.0.
• Allow plot_chroms and plot_chroms_heatmap to accept peak table as first argument.

Bug fixes and other minor changes

• Corrected two minor bugs in get_purity: one which counted a pure timepoint as impure when the estimated noise variance was zero, and another which returned NA without warning when only one timepoint fell below the noise threshold.
• Fixed preprocess function incorrectly truncating chromatograms when outliers are present and dim1 is not specified.
• Fixed input validation in preprocess so lists containing non-matrix elements no longer pass validation without error.
• Fixed filter_peaktable and filter_peaks: rts with a non-zero tol matched every peak within the tolerance rather than the nearest one, so a single requested time could select most of the table. Also, criteria are now inclusive (min_rt = 5 retains rt == 5), and peaks with NA in some samples are no longer excluded by min_value.
• Added error when duplicated sample names are provided to get_peaktable or get_peaks since this will cause downstream issues.
• Fixed issue with visual tests due to new version of Kaleido.
• Added additional arguments as placeholder for subset.peak_table.
• Fixed bug in normalize_data causing loss of sample names when normalizing chromatograms.
• Formalized requirement for R version 4.1.0.
• Fixed plot_spectrum plots with plotly engine so they always maintain their axis labels.
• Standardized axis labels across plot_spectrum plots.
• Standardized titles of trace plots across engines so they always use unicode lambda character.
• Fixed reshape_chroms so it returns a factor for the sample column.
• Fixed plot_chroms so chromatograms are black when plotting a single chromatogram with plotly or ggplot.
• Fixed plot_all_spectra to consistently export spectra in wide format regardless of plotting engine.
• Fixed error in equation describing spectral clustering in documentation for get_peaktable.
• Fixed get_peaktable (with clust = "sp.rt") to no longer take the absolute value of spectral similarity scores, which caused negatively correlated spectra to be treated as similar.
• Fixed bug in FWHM calculation in get_peaktable for raw peak shape.
• Suppressed spurious “essentially perfect fit” warning from summary.lm when computing R² in fit_peaks.
• Fixed subset.peak_table(), where sample_meta was subset on the wrong condition.
• Made substantial revisions to documentation to improve clarity.

chromatographR 0.7.5

• Fixed bug in check_chr when multiple chromatograms are provided.

chromatographR 0.7.4

• Added new heatmap plot in plot_chroms for plotting a list of chromatograms as a heatmap.
• Added labeling of spectral peaks for ggplot2 and plotly plots in plot_spectrum.
• Added minimum amplitude for labeling of spectral peaks in plot_spectrum, such that peaks are given a text label only if they exceed 5% of the maximum signal intensity.
• Removed parallel and mc.cores argument from preprocess (deprecated in v0.5.0).
• Added title argument in plot_chroms.
• Added option for different plotting engines through engine argument in plot.ptw_list.
• Added option for heatmap plots through what argument in plot.ptw_list.
• Changed behavior of reshape_chroms such that it can be applied to 2D chromatograms without throwing an error.
• Added index of functions to the package landing page.
• Made minor updates to function documentation.

Updates to GC-FID vignette

• Changed alignment plots to heatmaps rather than traces.
• Hid ggordiplot function in “folded” code block
• Added alt text to images.
• Updated call to adonis2 in accordance with new defaults in vegan v2.7-0.
• Made other minor changes to text.

Updates to chromatographR vignette

• Added figure demonstrating comparison of alignments with new plot_chroms_heatmap function.
• Added explanatory figure captions.

chromatographR 0.7.3

• Updated reshape_peaktable function to include wavelength and retention time data.
• Updated pk_tab data file to latest peak table format.
• Fixed minor issues with roxygen formatting.
• Added functional “families”.
• Fixed bug in get_peaks causing occasional errors due to edge cases.
• Added titles for ggplot2 and plotly plot_spectrum plots.
• Fixed bug in console output when providing chromatogram index as string to plot_spectrum.
• Changed default ggplot2 theme to theme_light.

chromatographR 0.7.2

• Export get_times and get_lambdas functions.
• Small revisions to documentation.
• Use tryCatch to allow for missing spectra in plot_all_spectra.
• Fixed rounding of p-values in cluster_spectra.
• Changed alpha in cluster_peaks to match common usage, such that the alpha parameter now specifies the significance level rather than the confidence level (1-alpha).
• Deprecated peak_no argument in cluster_peaks in favor of new min_size and max_size arguments.

chromatographR 0.7.1

• Fixed bug in get_peaks causing peaks to erroneously filtered out in some cases.
• Made small updates to documentation (in preprocess and fit_peaks functions) to better describe arguments.
• Added warning in mirror_plot when var contains more than two levels and levels aren’t specified.
• Fixed bug to allow mirror_plot to work properly with 2D data.
• Return dimnames for 1D ptw model objects returned by correct_rt.
• Minor updates to vignette.

chromatographR 0.7.0

• Updated correct_peaks function so it works properly for correcting retention times in peak lists.
• Added fixed_levels argument to reshape_peaktable so features can be plotted in the order they’re provided by the user.
• Added option for summing split peaks using the merge_peaks function by selecting method = "sum".
• Updated get_peaktable so that the use.cor argument works correctly (to use corrected retention times stored in a separate column).
• Fixed mirror_plot so it can take numeric input for lambdas.
• Changed default setting of verbose argument in correct_rt from FALSE to default setting.
• Removed load_chroms function. Use read_chroms instead.
• Eliminated spurious warning from attach_ref_spectra function.
• Changed name of index argument in plot.peak_list to idx. The original argument is now deprecated.
• Deprecated chr argument in multiple plotting functions (e.g., plot_spectrum) in favor of new idx argument.
• Fixed bug affecting plot_purity argument in plot.peak_list.
• Fixed bug in reshape_chroms so empty metadata column no longer appears.
• The plot_spectrum function now includes the peak names when plotting spectra.
• Fixed correct_rt so it no longer requires user-provided lambdas for 1D chromatograms.
• Added subset.peak_table function for easily subsetting peak_tables (e.g. to exclude specific peaks or samples).
• Added what argument for plot_all_spectra (e.g. to plot multiple spectra at a particular retention time).

Refactoring of cluster_spectra function:

• For simplicity, cluster_spectra now requires reference spectra to be attached to peak table.
• Accordingly, the chrom_list argument is no longer needed.
• Saving to RDS is now turned off by default.
• The pvclust package is now suggested instead of being required.

Updates to vignette and documentation

• Suggest numeric input to lambdas instead of character input to reduce unnecessary confusion.
• Made other minor changes to text of vignette to (hopefully) improve clarity.
• Added a short section on the attachment of reference spectra.

chromatographR 0.6.1

• Fixed bug in plot functions (e.g. plot_chroms and plot_spectrum) causing error when retention times are inconsistent between chromatograms.
• Eliminated spurious warning in preprocess function.
• Updated read_chroms syntax in vignette.

chromatographR 0.6.0

New features

• Enabled use of parallel package for parallel processing (in addition to current options using mcapply). (These options require the installation of suggested package pbapply).
• Updated get_peaktable for greater flexibility (e.g. for usage of ‘ChemStation’ peak lists as input).

Other changes

• Made some minor changes to vignette to improve clarity (e.g. using single wavelength for integration, etc.)

Bug fixes

• Fixed error in attach_metadata when there are NA values in merge column.

chromatographR 0.5.6

• Fixed bug in preprocess function causing fatal error due to misrecognition of matrices.
• Fixed behavior of plot_chroms and correct_rt to allow automatic detection of lambda for 1D chromatograms.
• Fixed bug in combine_peaks (due to misplaced parenthesis).
• Added new option to filter by maximum peak area or height in filter_peaktable (what = "max"), as suggested by Katherine Holmes.

chromatographR 0.5.5

• Fixed bug in get_peaktable causing failure to print strip plot when plot_it == TRUE.

chromatographR 0.5.4

• Added .zenodo.json file.

chromatographR 0.5.3

• Fixed bug in plot_chroms causing mismatched legend labels in base R plot.
• Added additional arguments to plot_chroms function: xlim,ylim, and legend_position.
• Added additional information about arguments available in get_peaks for fine-tuning the peak-finding algorithm (in response to #27).

chromatographR 0.5.2

• Added metadata argument to reshape_peaktable for filtering metadata fields.
• Added option for renaming peaks via reshape_peaktable by providing a named character vector.

chromatographR 0.5.1

Bug fixes

• Fixed syntactical bug in get_peaktable when applied to gaussian peak list.
• Fixed bug caused by improper transfer of time.units metadata by filter_peaks function.
• Added default for missing time.units in plot.peak_list.

chromatographR 0.5.0

New features

• Added ggplot2 option to plot_spectrum, plot.peak_table and plot_all_spectra functions.
• Reworked write_chroms for more sensible handling of paths and added filename argument.
• Updated get_purity function to improve speed.
• Added additional argument to reshape_chroms function for subsetting data by retention times (rts).
• Added parallel processing through the pbapply package for the correct_rt, get_peaks, and preprocess functions by setting the cl argument.

Other changes

• Changed behavior of preprocess when inferring retention times so chromatograms are no longer rounded down to the largest integer.
• In preprocess, spectral smoothing is no longer applied on 2D chromatograms, removing error message when preprocess is used with default settings.
• Moved position of ... argument to end in plot.peak_table.
• Changed progress_bar argument to show_progress in correct_rt, preprocess and get_peaks to fix strange pmatch behavior with additional arguments to preprocess.
• Changed orientation of “plotly” plots generated by plot_spectrum to match other plotting engines.
• Deprecated the mc.cores argument in correct_rt is now deprecated in favor of the new cl argument.
• Deprecated the parallel argument in preprocess in favor of just using cl.
• Changed name of first argument in mirror_plot from peak_table to x. Otherwise the function has not changed.
• Added additional tests, improving test coverage to 80%.
• Updated get_chrom_list (internal) to allow parsing of subsetted lists.

chromatographR 0.4.8

• Fixed bug in merge_peaks function so it works properly (to combine 2 or more peaks in a peak table).
• Fixed bugs in plot_chroms preventing plotting with ggplot2 and plotting wrong chromatograms in base R.
• Added additional tests for plot_chroms and reshape functions.

chromatographR 0.4.7

• Added reshape_peaktable function for conversion of peak tables to long format.
• Turned off estimate_purity in get_peaks function by default.
• Added option to filter by wavelength in reshape_chroms, speeding up plot_chroms.

chromatographR 0.4.6

New Features

• Added plot_chroms function for easily plotting multiple traces from a list of chromatograms.
• Minor changes to internal syntax of correct_rt to give more informative error messages.
• Added estimate_purity argument in get_peaks to toggle peak purity estimation.
• Changed default setting for progress_bar in correct_rt and get_peaks. Now defaults to TRUE if pbapply is installed.
• Added additional tests of utility functions and new plot_chroms function.
• Minor changes to vignette.
• Minor changes to documentation.

Bug fixes

• Fixed bug causing mismatched time axes and alignment issues after VPdtw warping (again), so that it returns matrices with a consistent time axis.
• Fixed y unit label in boxplot.peak_table function.
• Fixed behavior of plot_spectrum so spectrum is exported properly when engine == plotly.
• Fixed bug in write_peaktable when writing to xlsx.

chromatographR 0.4.5

New Features

• Added reshape_chroms function for converting chromatograms to “long” format.
• Added write_peaktable function to easily write peak_table to csv or xlsx.
• Added get_purity function for assessing peak purity.
• Allow multiple peaks as arguments to plot.peaktable.
• Added functions for plotting traces and spectra with plotly: plotly_trace and plotly_spec.
• Fixed preprocess so it will no longer try to interpolate along columns for 2D data.
• Added stand-alone boxplot function for peak_table objects.
• Added a new class (ptw_list) and plotting function for lists of ptw alignment objects.
• Added plot_it argument in correct_rt for plotting alignments.
• Added VPdtw as a dependency (instead of being only suggested).
• Added progress_bar option to get_peaks and correct_rt.
• Improved error handling in plot.peaklist.
• Updated find_peaks function with more and better smoothing options to improve peak-finding. Now defaults to gaussian smoothing.
• Changed fit_peaks function and syntax (see below).
• Minor updates to vignette.

Changes to fit_peaks function:

• Simplified logic in fit_peaks function.
• Modified fit_peaks syntax so it now takes a matrix (x) and a wavelength (lambda) instead of a numeric vector (y).
• Incorporated assessment of peak purity during peak fitting.
• Added wavelength (lambda) to peak_list and peak_table metadata.
• Fixed bug to allow fitting of a single peak with fit_peaks.

Bug fixes

• Fixed bug causing mismatched time axes (and improper alignment of chromatograms) after variable penalty dynamic time warping (VPdtw).
• Fixed bug in attach_metadata that could result in disordered rows.
• Fixed occasional test failure on MKL server by skipping cluster_spectra test on CRAN.
• Adjusted cluster_spectra and combine_peaks functions so messages can be suppressed with verbose == FALSE.

chromatographR 0.4.4

CRAN release: 2022-08-24

• Fixed issue with tests when run on certain machines (MKL).

chromatographR 0.4.3

CRAN release: 2022-08-22

• Minor changes to documentation.
• Added additional check of chrom_list dimensions and names.

chromatographR 0.4.2

New features

• Added option to select time.units for peak area in get_peaks function facilitating better comparison with vendor software.
• Now allow preservation of instrumental metadata through pre-processing and alignment steps.
• Added filter_peaktable function.

Minor changes:

• Deprecated load_chroms function. Please use read_chroms from chromConverter to import files instead.
• Changed default behavior in correct_rt to corrected_values rather than models.
• Added more informative warnings and error messages to various functions.
• Now recommend installation of VPdtw from CRAN instead of https://ethanbass.github.io/drat/
• Fixed typos in vignette

Bug fixes

• Fixed bug in cluster_spectra affecting peaks with 0 standard deviation.
• Fixed bug affecting peak_list metadata.

chromatographR 0.4.1

CRAN release: 2022-05-19

• Extended package DESCRIPTION and added citations to relevant references.
• Added \value and \section{Side effects} fields to docs for the various plot functions.
• Fixed bug in mirror_plot so legend can be fully hidden by setting plot_legend to FALSE.
• Fixed bug in scan_chrom so additional arguments are passed to plot_spectrum.
• Added color argument to customize color of fitted peaks in plot.peak_list.
• Fixed plot functions and examples so they don’t change par settings.
• Other small updates to documents (mostly formatting or small clarifications).
• Fixed plot.peak_table so it can return spectra if export_spectrum is TRUE.
• Added error for box_plot option in plot.peak_table if metadata is not attached.
• Added error for box_plot option in plot.peak_table if peak is not provided to loc.
• Added error in plot_spectrum function for user-supplied retention times beyond the edges of the chromatogram.
• Added error in plot_spectrum function for unspecified lambda (if what=="click).
• Released on CRAN

chromatographR 0.4.0

• Added support for variable penalty dynamic time warping (VPdtw) through correct_rt function.
• Fixed bug in get_peaks function.
• Allow preprocessing without interpolation.
• Fixed bug so preprocess can work on Windows (without parallel processing).
• Allow use of raw data for peak integration in get_peaks.
• Added verbose option to correct_rt to print reference chromatogram.

chromatographR 0.3.0

• Added a NEWS.md file to track changes to the package.