Fast and flexible processing of large FRET image stacks using the FRET-IBRA toolkit.

Ratiometric time-lapse FRET analysis requires a robust and accurate processing pipeline to eliminate bias in intensity measurements on fluorescent images before further quantitative analysis can be conducted. This level of robustness can only be achieved by supplementing automated tools with built-in flexibility for manual ad-hoc adjustments. FRET-IBRA is a modular and fully parallelized configuration file-based tool written in Python. It simplifies the FRET processing pipeline to achieve accurate, registered, and unified ratio image stacks. The flexibility of this tool to handle discontinuous image frame sequences with tailored configuration parameters further streamlines the processing of outliers and time-varying effects in the original microscopy images. FRET-IBRA offers cluster-based channel background subtraction, photobleaching correction, and ratio image construction in an all-in-one solution without the need for multiple applications, image format conversions, and/or plug-ins. The package accepts a variety of input formats and outputs TIFF image stacks along with performance measures to detect both the quality and failure of the background subtraction algorithm on a per frame basis. Furthermore, FRET-IBRA outputs images with superior signal-to-noise ratio and accuracy in comparison to existing background subtraction solutions, whilst maintaining a fast runtime. We have used the FRET-IBRA package extensively to quantify the spatial distribution of calcium ions during pollen tube growth under mechanical constraints. Benchmarks against existing tools clearly demonstrate the need for FRET-IBRA in extracting reliable insights from FRET microscopy images of dynamic physiological processes at high spatial and temporal resolution. The source code for Linux and Mac operating systems is released under the BSD license and, along with installation instructions, test images, example configuration files, and a step-by-step tutorial, is freely available at github.com/gmunglani/fret-ibra.

This is a PLOS Computational Biology Software paper.

Introduction

Ratiometric biosensors based on FRET (Förster Resonance Energy Transfer) are often used to quantify the dynamics of physiological processes at subcellular resolution. Several methodologies have been developed to optimize the calibration of these protocols [1, 2], but the amount of data generated still requires fast, scalable, and flexible software tools for efficient and robust analyses.

Accurate retrospective background subtraction and photobleaching correction algorithms are critical for the extraction of precise spatio-temporal pixel intensity distributions from time-lapse images. This is particularly true for studies relying on the time series analysis of ion concentrations linked with cell kinematics [3]. The inherent variability of a single experiment over longer time scales necessitates the use of heuristically defined process parameters that can be applied to specific data splits rather than the entire image stack to combat temporal drift [4].

Several algorithms have been implemented as stand-alone packages or within existing tools (ImageJ, CellProfiler, and qTfy) to facilitate the efficient processing of a variety of microscopic imaging modalities. Traditional methods are typically very robust and widely applicable but often use a static process parameter set, which is not trivial to customize for specialized applications [5-7]. A number of high-performance algorithms designed for retrospective background subtraction in light microscopy images, using clustering and constrained minimization methods, have been developed to improve precision. However, they have been shown to be generally unsuitable for characterizing growing cells by time-lapse images, which exhibit high temporal correlation [4, 8–10].

In this work, we describe FRET-IBRA (Image Background-subtracted Ratiometric Analysis), a fully parallelized, configuration file-based tool developed to simplify the ratiometric analysis of FRET image stacks [11]. FRET-IBRA accepts multi-image TIFF stacks with different bit-depths (8, 12, and 16 bit) as input, and outputs both multi-image TIFF and HDF5 stacks for further downstream analyses. To provide maximum flexibility, this tool incorporates the processing of discontinuous frames within an image stack as well as the correction of individual frames with optimal parameters to ensure that temporal noise is minimized. It should be noted that the ratiometric approach described here is limited to linked constructs of fluorescent proteins where pixelwise stoichiometry is maintained. To showcase the performance of this tool, FRET images of growing Arabidopsis thaliana pollen tubes are processed to create ratio stacks to evaluate the spatial distribution of calcium ions driving growth.

Design and implementation

Description

FRET-IBRA consists of three modules which are responsible for (i) the background subtraction of acceptor and donor time-lapse images, (ii) the optional registration and alignment of these processed image stacks in case a dual-view setup was used to produce ratio images, and (iii) the photobleaching correction, respectively (Fig 1A). Furthermore, the tool outputs several metrics to gauge the performance of the background subtraction algorithm on individual frames, allowing for the fast detection and reprocessing of sub-optimally processed frames. Logging and HDF5 storage capabilities are also provided to aid in reproducibility.

Fig 1

FRET-IBRA workflow.

A) Image evolution from the original image to the background-corrected image and the final ratiometric image. B) A contour map of the pixel intensities of the original and corrected images. C) Pixel intensity distributions of example background and foreground tiles shown in B), along with the reduced feature space showing median, skewness, and variance, with each point representing a tile. The blue cluster indicates the background tiles, while the red cluster indicates foreground tiles. D) The percentage of foreground pixels can be used to control the quality of the background subtraction for each channel. E) The median signal intensity of donor and acceptor channel allows for analysis and correction of bleaching effects, and serves as a an additional quality indicator for background subtraction.

Modules

The background subtraction module tiles the image frame into squares of a set dimension and uses the DBSCAN clustering algorithm with an Euclidean metric to identify tiles as foreground or background (Fig 1B and 1C). This algorithm is adapted from Schwarzfischer and colleagues (2011) [8], using an expanded feature space, which includes the median and higher moments (standard deviation, skewness, and kurtosis) of the tiles’ pixel intensity distribution along with the position of its intensity-weighted centroid (Fig 1C). The background tile intensities are used to quadratically interpolate over the entire grid to estimate the spatially varying background intensity caused by shading effects. The resulting background intensity is then subtracted from the original image (Fig 1B). This method requires that, for the DBSCAN algorithm, a tuning parameter ϵ is provided in the configuration file. ϵ is defined as the maximum distance, in feature space, between two tiles to be classified as the same cluster. This algorithm is effective in correcting spatial shading and removing local noise along with providing accurate background subtraction for images with few objects of interest, but is optimized for single-cell images. When dealing with images with very low signal intensities, users should carefully inspect the corrected images before proceeding as very low signals can produce artefacts (brachypodium_roots example in the FRET-IBRA repository). This module outputs an animation of frame-by-frame metrics of algorithm performance for visual inspection as well as the filtered HDF5 dataset and TIFF image stacks.

The ratiometric processing module contains several optional pre-processing steps before the ratio image is produced. Image registration between the donor and acceptor stacks is first performed by rigid linear transformation, followed by the application of a small kernel bilateral smoothing filter. Otsu’s thresholding is then used to binarize the images, after which singularities at the image boundary are removed. The percentage of non-zero (foreground) pixels per frame of the donor and acceptor stacks (Fig 1D) and the ratio of median intensity to bit depth (Fig 1E) are provided as a quantitative metric of the algorithm’s performance over the entire image stack. Sharp discontinuities in these metrics between sequential frames may indicate that specific frames require reprocessing with a different ϵ. The ratiometric images, which represent the ratio of the acceptor and donor image stacks, are saved as a HDF5 dataset and a TIFF image stack.

Photobleaching correction is then optionally performed on frames of the acceptor and/or donor stacks, using a regularized linear or exponential fit on the median frame intensity (Fig 1E). The frame range used for the fit is specified in the configuration file and applied to all frames starting with the onset of photobleaching, defined as the lower bound of the provided frame range. The corrected ratiometric images are saved as an output TIFF image stack, while the bleach correction factors are stored in the HDF5 dataset for further analysis, if required.

FRET-IBRA is built to process images with few objects of interest that are densely packed, allowing for accurate background estimation. As the background subtraction algorithm classifies tiles solely as foreground or background, images with a wide spatially homogenous distribution of cells or other objects of interest might not be processed correctly due to poor estimation of the background signal.

Results

To showcase the efficacy of FRET-IBRA, FRET images displaying the calcium distribution in growing Arabidopsis thaliana pollen tubes were corrected with the toolkit and the result was compared with existing background subtraction tools like BaSiC [10] and the background subtractor module of the Mosaic image processing package [12] (Fig 2A). The optimal background window size parameter in FRET-IBRA, which determines how many square tiles the image width should be divided into, is determined to be 40 for this image stack. This parameter requires tuning based on the image feature dimensions, to ensure accurate foreground extraction without utilizing excessive computational resources. Furthermore, it should be emphasized that the higher the number of tiles, the smaller the tile size and the longer the runtime of the background subtraction algorithm.

Fig 2

Comparison of FRET-IBRA with existing background subtraction plugins for ImageJ (BaSiC [10] and the background subtractor module of the Mosaic suite [12]).

A) Frame 1 of the donor (CFP) channel after background subtraction. The dashed lines indicate the rows of pixels analysed in B). B) Pixel intensities after background subtraction. Top and bottom panels correspond to lines 1 and 2, respectively. C) Runtime for the background subtraction for 1000 frames. All tests were run on 4 cores with an Intel Core i7 (Quad-Core, 2.8GHz, 16GB RAM) computer.

As can be seen by two line samples of the image pixel intensities in Fig 2B, FRET-IBRA is the most effective package for subtracting background pixel intensities whilst maintaining the peak intensity value of the pollen tube. Furthermore, in addition to this background subtraction, FRET-IBRA smoothens the resulting background pixel intensities by reducing the standard deviation of the background noise distribution. With 4 cores, the runtime for FRET-IBRA lies between BaSiC and Mosaic (Fig 2C). However, unlike the other packages, the parallel implementation of FRET-IBRA allows for further runtime reduction that scales linearly with an increasing number of cores.

The BaSiC tool is shown to be effective in shading correction of the image when compared to the original; however, the average background pixel intensity remains mostly unaltered (Fig 3). In contrast, the background subtractor module of Mosaic successfully subtracts the intensity of the background pixels, although the standard deviation of the resulting background noise remains relatively high (Table 1). While the window size for Mosaic was set at 35, it should be noted that this parameter does not seem to have a significant effect on the accuracy of this tool.

Fig 3

Background subtraction comparison.

Background levels according to dashed line 2 in Fig 2 after background subtraction with FRET-IBRA, BaSiC, and Mosaic compared with the original image. The BaSiC plugin corrected only the shading, while Mosaic and FRET-IBRA corrected the shading and subtracted the background. Overall, processing the image with FRET-IBRA results in the background with the lowest pixel intensity and standard deviation.

Table 1

Pixel intensity statistics after running background subtraction algorithm with variable window sizes.

The window size in FRET-IBRA defines the number of tiles the image width should be divided into, i.e., for a 640x480 pixel image, a window size set at 40 results in 40 windows along the width and 30 windows along the height (1200 windows with a tile size of 16x16 pixels). The window size cannot be set for BaSiC. Parameters used for our comparison are marked yellow.

Pixel Intensity	FRET-IBRA			Mosaic			BaSiC
Window	20	40	80	20	35	70	-
Mean	5.35	4.92	5.33	8.41	8.24	8.73	275.21
Max	41	27	10	51	50	50	320
Min	1	2	4	0	0	0	228

Ratiometric image stacks were then produced for FRET-IBRA, BaSiC, and Mosaic from the processed acceptor and donor stacks (Fig 4). The resultant ratio images were scaled to display the full 8-bit range, and the first frame processed by each tool was used for comparative purposes (Fig 4A). The background signal in the BaSiC-processed ratio image was uniformly high, which generally allowed for good discrimination between foreground and background signal, with the exception of a lower intensity halo surrounding the pollen tube outline (Fig 4A, top panel). Visually, the dynamics (signal spectrum) of the foreground can be seen to suffer from high background values, rendering any quantitative analysis very difficult. The ratio image produced by Mosaic was significantly better due to the resulting high signal to noise ratio of the input images (Fig 4A, middle panel). The acceptable background intensity, however, was diminished by irregular noise with occasional high intensity peaks that could not be completely eliminated with additional filtering, along with the continued presence of a halo around the pollen tube outline. In contrast to BaSiC and Mosaic, ratio images produced after FRET-IBRA background subtraction displayed a low uniform background signal, leading to a superior signal-to-noise ratio without strong artifacts and extreme outliers (Fig 4A, bottom panel).

Fig 4

Ratio image processing.

A) Background-subtracted acceptor and donor images from BaSiC-, Mosaic-, and FRET-IBRA-produced ratio images were processed and compared. Surface plots created from the ratio images give a better overview of the background subtraction quality and the signal-to-noise ratio achieved with each software (top: BaSiC, middle: Mosaic, bottom: FRET-IBRA). B) FRET-IBRA automatically registers and rescales the ratio image with robust rescaling to reveal the entire range of the signal spectrum.

From the FRET-IBRA processed image in Fig 4A, it is evident that mapping the full 8-bit signal intensity range is not ideal for visually representing the dynamics of the displayed calcium levels. The reason is that even a single outlier with an unexpectedly high value would determine the scale factor of the entire image. The equivalent also occurs at the lower end of the foreground signal range, leading to a relatively narrow intensity histogram and a flat image. Therefore, FRET-IBRA uses robust rescaling (10th—90th percentile) on the ratio image intensity values to reveal the hidden dynamics of the signal intensity. The registered and rescaled version of the ratio image clearly showcases a high distribution of calcium ions at the tip with a sharp drop-off outside the tip region. As expected, the informative quality of a ratio image hinges largely on achieving a robust signal-to-noise ratio from the input stacks, which is, in turn, highly dependent on the background subtraction algorithm.

FRET-IBRA is a fully-parallelized, modular, and flexible tool that provides a complete workflow solution to conduct ratiometric analyses of FRET images. The package has been shown to correct shading, effectively subtract background pixel intensities, and smoothen background noise more effectively than comparable packages for images with few objects of interest. In addition, FRET-IBRA creates ratiometric images and performs photobleaching correction in a modular fashion, allowing for efficient parameter tuning for large image stacks. It should be noted that bleaching correction must be used with caution as it can cause artifacts, such as masking of true signal intensity variations or overcorrection, resulting in a false increase in signal intensity if the detrending methods provided are not appropriate. Therefore, it may sometimes be advisable to accept overall attenuation of the resulting ratio signal in order to detect short-term fluctuations. Alternatively, bleaching correction can be performed using other detrending algorithms outside of FRET-IBRA. The resultant images produced by FRET-IBRA on FRET images of growing pollen tubes have been shown to have high signal-to-noise ratios and no significant artifacts, allowing for accurate further quantitative or qualitative downstream analyses. Furthermore, the flexible nature of FRET-IBRA enables discontinuous frames to be easily processed and corrected individually, resulting in potentially significant time savings. FRET-IBRA performs well with various fluorophores tested, including biosensor proteins and fluorescent dyes. Background subtraction can also be performed using single stacks.

Future directions

While FRET-IBRA is primarily built to be a command-line tool using configuration file-based parameters, the classes implemented in Python3 are modular and can be easily integrated into a larger analysis pipeline. Further developments are currently under way to locate regions of high ratiometric intensity and track their time-varying dynamics.

25 Aug 2021

Dear Dr. Vogler,

Thank you very much for submitting your manuscript "Fast and flexible processing of large FRET image stacks using the FRET-IBRA toolkit" for consideration at PLOS Computational Biology.

As with all papers reviewed by the journal, your manuscript was reviewed by members of the editorial board and by several independent reviewers. In light of the reviews (below this email), we would like to invite the resubmission of a significantly-revised version that takes into account the reviewers' comments.

We cannot make any decision about publication until we have seen the revised manuscript and your response to the reviewers' comments. Your revised manuscript is also likely to be sent to reviewers for further evaluation.

When you are ready to resubmit, please upload the following:

[1] A letter containing a detailed list of your responses to the review comments and a description of the changes you have made in the manuscript. Please note while forming your response, if your article is accepted, you may have the opportunity to make the peer review history publicly available. The record will include editor decision letters (with reviews) and your responses to reviewer comments. If eligible, we will contact you to opt in or out.

[2] Two versions of the revised manuscript: one with either highlights or tracked changes denoting where the text has been changed; the other a clean version (uploaded as the manuscript file).

Important additional instructions are given below your reviewer comments.

Please prepare and submit your revised manuscript within 60 days. If you anticipate any delay, please let us know the expected resubmission date by replying to this email. Please note that revised manuscripts received after the 60-day due date may require evaluation and peer review similar to newly submitted manuscripts.

Thank you again for your submission. We hope that our editorial process has been constructive so far, and we welcome your feedback at any time. Please don't hesitate to contact us if you have any questions or comments.

Sincerely,

Dina Schneidman-Duhovny

Software Editor

PLOS Computational Biology

Dina Schneidman-Duhovny

Software Editor

PLOS Computational Biology

***********************

Reviewer's Responses to Questions

Comments to the Authors:

Please note here if the review is uploaded as an attachment.

Reviewer #1: The data analysis pipeline presented in “Fast and flexible processing of large FRET image stacks using the FRET-IBRA toolkit” tackles major issues in obtaining robust quantitative estimates from fluorescence microscopy time series, especially when dealing with ratiometric reporters. The authors employ solid image processing techniques that can: a) subtract the background for each frame even with spatial variation; b) perform photobleaching correction; c) output a scaled ratio image that could be used for further quantitative analysis. This open-source tool has a clear and flexible methodology built on freely available and well-written Python code, which should appeal directly to the research of single cell growth/migration.

FRET-IBRA has the potential to become the go-to tool for researchers in the pollen tube and root hair community analyzing ratiometric data, which I believe would be enhanced if the authors address issues on three fronts: 1) benchmark; 2) automation; 3) distribution/interface. Although FRET-IBRA seems to offer a superior output than other related tools, this comparison was made with a single illustrative example and could benefit from directly quantifying variables of biological interest across multiple data-sets. Furthermore, automating the choice of certain parameters (mainly epsilon) could offer more objective criteria and minimize manual intervention. Lastly, despite providing a clear code and installation instructions in GitHub, the current distribution and interface requires a moderate level of programming experience that limits the potential users.

Regardless of the potential shortcomings, FRET-IBRA is certainly a useful tool that I personally intend to incorporate in my analyses. Below I further discuss points that can, hopefully, increase the adoption of this pipeline.

MAJOR POINTS

1) Benchmark

The authors use an example dataset to illustrate the workflow of FRET-IBRA and to compare it with other tools. While the representative dataset did provide a clear explanation of the method and its performance, it is difficult to evaluate how generalizable are the results when dealing with different datasets. Does the difference in performance with other tools hold up with different fluorescent reporters, microscope settings, image and object sizes? Relying on manual choice of parameters further complicates this matter (see discussion on automation), which is not to say that a manual option is not welcome. Furthermore, since the aim is to provide robust quantitative imaging, ideally one should also compare estimates of quantities-of-interest such as tip fluorescence ratio and tip-to-shank gradient. Showing the potential of FRET-IBRA to provide new biological insights from reliable spatiotemporal quantification of ratiometric fluorescence would greatly increase its impact within the aims of PLOS Computational Biology.

2) Automation

Although the possibility of correcting individual frames may be interesting when discontinuous events occur, requiring visual inspection and manual parameter choices compromises objectivity and increases the time required to analyze each series. How could one ensure comparable results when drastically different parameters are used to process different frames, channels and series?

I encourage the authors to provide an automatic choice of epsilon (and maybe even window size), which even if just “guesstimates” may help users to tune them as little as possible. Perhaps a few simplifying assumptions can aid in that endeavor, for example assuming that the number and location of foreground and background tiles should be very similar (if not identical) between channels.

Finally, while I fully appreciate that some experiments may lead to outliers, overcorrecting for them may lead to greater artifacts than simply disregarding specific time points. On that note, I would like to understand more the reasoning behind using image registration between acceptor and donor channels. Unless both channels are sampled with a large time interval, it seems to me that they should be nearly or completely identical. Thus, allowing for image registration – even if only a rigid linear transformation – seems to increase the chance of introducing artifacts. Although the online tutorial does mention that registration should be used only after choosing an “optimal” value of epsilon, there is seemingly no objective criteria for that. I can picture scenarios where an experimental intervention may displace the cell, where registration would be important to match its position, however this would not be between channels but between consecutive frames. I ask the authors to consider adding further explanation or a discussion on the matter.

3) Distribution/interface

The choice of implementing the FRET-IBRA pipeline in Python has multiple benefits but also imposes challenges to users without a moderate programming experience. Even for initiates, some of the requirements of the pipeline are notoriously cumbersome to install. Despite the clear GitHub instructions, in my personal experience the installation took considerable time and effort. Even after managing the installation, there were errors in the ratiometric module and in processing 1024x1024 images that I am unsure whether they stem from issues in the code or simply installation flaws (see attached report). Two of the datasets tested yielded apparently great results in terms of background subtraction in both channels, however the ratiometric module yielded the error: “IndexError: cannot do a non-empty take from an empty axes” (see report). The other two datasets with 1024x1024 images were not processed due to a “ValueError: cannot reshape array of size 1048576 into shape (40,25,newaxis,25)” (see report).

Regardless of my personal experience testing the pipeline, a few solutions would definitely attract more users. The most inclusive option would be a web-interface where the user could interactively tune parameters. A Python-based graphical user interface could also help attract users, but it would likely still require cumbersome installations. An interactive Jupyter notebook can be a great choice but still may require a little programming skill. I encourage the authors to consider what is their audience and what would be the friendliest (and yet achievable) solution.

Finally, if sufficient automation is achieved, it would be interesting to have the complete workflow performed sequentially for a given series with a single command (background subtraction for each series, ratiometric calculation and optional photobleaching correction). Ideally a “batch mode” could ensure that a complete data cohort is analyzed with comparable parameters, although it may still be desirable to maintain a level of intervention in specific frames.

MINOR POINTS

- Lacking a discussion on how correcting photobleaching can insert artifacts.

- The standard deviation can be arbitrarily reduced with the amount of smoothing used (Pg. 5 ln 112-113). This maybe the reason why we see a decrease in the standard deviation and maximum values with window size in table 1. Thus, it maybe not be completely fair to use the standard deviation to compare with methods that do not perform smoothing.

- Pg. 6 ln 132 “rendering any quantitative analysis impossible” is a strong assertion, I recommend softening it.

- Pg. 6 ln 146-147 “true dynamics of the signal intensity” is a strong assertion, I recommend substituting “true” by “hidden” or other equivalent term.

- Pg. 7 ln 150 “best possible signal-to-noise ratio” is a strong assertion, I recommend softening it.

Reviewer #2: Summary:

The paper proposes a processing tool for FRET image stacks, including i) background subtraction for acceptor and donor time-lapse images, ii) registration between acceptor and donor movies to produce ratio images and iii) photobleaching correction. It can be used to quantify the spatial distribution of calcium ions during pollen tube growth under mechanical constraints and the code is released under BSD license.

Generally, it is not a methodologically innovative paper, as the primary methods involved in the tool are based on existing methods. For example, background subtraction is based on DBSCAN clustering proposed in another paper. Registration between acceptor and donor movies are a standard rigid-body registration. Nevertheless, a tool that comprises multiple processing steps could be still useful in practice is if its usability and robustness are sufficiently demonstrated. Hence, I would suggest authors should consider evaluating their tool more thoroughly, use multiple movies, best captured by different microscopy under different conditions, or in multiple labs, to demonstrate the generalisability of the proposed tool. These issues should be addressed in the revision.

Major issues:

i) The background subtraction module is based on a DBSCAN clustering algorithm on image tiles to separate foreground and background, proposed in Schwarzfisher et al (2011). Since this clustering needs to be done for each frame in the movie and is generally slow, does the proposed tool improve the original algorithm in terms of speed?

ii) Comparison between BaSiC: as far as I know, BaSiC has a function to correct background (by setting Fiji plugin “Temporal_drift of baseline” to be “replace with zero” and it has a zero-like background by then. So it is surprising that your BaSiC corrected movie still has a high background value. Nevertheless, BaSiC might have a problem with highly-correlated foreground, which seems to be the case of the present example (exemplary movie in the github example) and may not function properly (as that breaks the basic assumption of BaSiC)

iii) How many movies are in the evaluation set. It seems to me only one pair of movies, i.e. acceptor and donor, is used. In my opinion, that is not sufficient to evaluate the robustness of the tool, which is also my main criticism of the paper.

Suggestions

i) If authors can also provide some biological-relevant down-stream analysis, they would be a more solid demonstration of the necessity of such a tool.

ii) Authors should also discuss the potential usage and limitation of the tool: when it works and when it could fail, to prevent misuse of the tool.

**********

Have the authors made all data and (if applicable) computational code underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data and code underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data and code should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data or code —e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: No: The code is fully available on GitHub with an example data set. However, the actual data set used to compare FRET-IBRA with other tools was not (probably due to size constrains).

Reviewer #2: Yes

**********

PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: Yes: Daniel Damineli

Reviewer #2: No

Figure Files:

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, . PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email us at .

Data Requirements:

Please note that, as a condition of publication, PLOS' data policy requires that you make available all data used to draw the conclusions outlined in your manuscript. Data must be deposited in an appropriate repository, included within the body of the manuscript, or uploaded as supporting information. This includes all numerical values that were used to generate graphs, histograms etc.. For an example in PLOS Biology see here: http://www.plosbiology.org/article/info%3Adoi%2F10.1371%2Fjournal.pbio.1001908#s5.

Reproducibility:

To enhance the reproducibility of your results, we recommend that you deposit your laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. Additionally, PLOS ONE offers an option to publish peer-reviewed clinical study protocols. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols

Submitted filename: Tests performed with FRET-IBRA.pdf

Click here for additional data file.

15 Dec 2021

Submitted filename: FRET-IBRA_response_to_reviewers.docx

Click here for additional data file.

2 Jan 2022

Dear Dr. Vogler,

Thank you very much for submitting your manuscript "Fast and flexible processing of large FRET image stacks using the FRET-IBRA toolkit" for consideration at PLOS Computational Biology. As with all papers reviewed by the journal, your manuscript was reviewed by members of the editorial board and by several independent reviewers. The reviewers appreciated the attention to an important topic. Based on the reviews, we are likely to accept this manuscript for publication, providing that you modify the manuscript according to the review recommendations.

Please prepare and submit your revised manuscript within 30 days. If you anticipate any delay, please let us know the expected resubmission date by replying to this email.

When you are ready to resubmit, please upload the following:

[1] A letter containing a detailed list of your responses to all review comments, and a description of the changes you have made in the manuscript. Please note while forming your response, if your article is accepted, you may have the opportunity to make the peer review history publicly available. The record will include editor decision letters (with reviews) and your responses to reviewer comments. If eligible, we will contact you to opt in or out

[2] Two versions of the revised manuscript: one with either highlights or tracked changes denoting where the text has been changed; the other a clean version (uploaded as the manuscript file).

Important additional instructions are given below your reviewer comments.

Thank you again for your submission to our journal. We hope that our editorial process has been constructive so far, and we welcome your feedback at any time. Please don't hesitate to contact us if you have any questions or comments.

Sincerely,

Dina Schneidman

Software Editor

PLOS Computational Biology

***********************

A link appears below if there are any accompanying review attachments. If you believe any reviews to be missing, please contact ploscompbiol@plos.org immediately:

[LINK]

Reviewer's Responses to Questions

Comments to the Authors:

Please note here if the review is uploaded as an attachment.

Reviewer #1: OVERALL REMARKS

Munglani et al. addressed several points raised by the reviewers and presented: 1) a friendlier interface to the FRET-IBRA pipeline both in terms of user input (GUI) and workflow (complete background subtraction on both channels and ratiometric processing); 2) more examples of its biological applicability in the GitHub repository and 3) improved presentation/discussion of the results. These changes enhance the usability of the tool, although there are still a few shortcomings in the presentation and discussion of the results pointed out by both reviewers in the previous considerations. In my opinion, the major limitation of the manuscript still is to provide a clear demonstration of the biological applicability of the pipeline, which should be easily achievable by including long ratiometric movies and time series. I suggest illustrating the potential of FRET-IBRA in reporting the known oscillations in cytosolic calcium concentrations at the pollen tube tip. I reckon such addition to the manuscript should be easily achievable with data from the authors themselves, otherwise I would be happy to provide a suitable data set.

DETAILED REMARKS

I am glad the authors took the effort to extend FRET-IBRA’s interface by providing a GUI and the option to do background subtraction for both channels and calculate the ratio, these changes should facilitate the accessibility of the pipeline for many users. Furthermore, the biological examples in the GitHub repository suggest a wide range of applicability of the tool although the data was not explicitly incorporated in the manuscript. Thus:

- Discussions about the examples in GitHub are limited since details are lacking

- Examples are single time points making it hard to evaluate the potential of the pipeline to generate a long ratiometric movie

- The example in Brachypodium roots suggest strong spatial artifacts introduced by the background subtraction, especially of the acceptor channel (see figure in the attached report)

- It is often difficult to interpret the biological relevance of the ratiometric output (e.g. Brachypodium roots example) without a clear reference, whereas studying its temporal behavior can provide greater reliability to the result

Regardless of other examples, the model system already presented in the manuscript (i.e. pollen tubes) is ideal to show whether the ratiometric processing can produce biologically meaningful results. Pollen tube growth is accompanied by notorious tip-focused oscillations in cytosolic calcium, which could be used as a proof-of-principle that the method can adequately capture biological oscillations.

MINOR POINTS

- I am still unable to produce ratiometric stacks using 16 bit images, even when using the suggested nwindow (or others I tried). Admittedly this can be some issue on my end or something extremely simple to fix (check log in the attached report)

- I suggest adding an option to use bg/fg tiles from one channel to the other when image registration is not being used. This makes sense given the pixelwise stoichiometry assumption and would cut processing time nearly in half while avoiding artifacts due to different bg/fg definition in acceptor and donor channels. I emphasize that this is not a mandatory suggestion.

- One idea is to include loess as a method for photobleaching correction, depending on the series length a sufficiently large span parameter should capture the slow trend. One idea is to use degree 1 only to avoid inserting artifacts in the trend. I emphasize that this is not a mandatory suggestion.

- Sentence in the abstract stating that the package has been “extensively used in quantifying the spatial distribution of calcium ions during pollen tube growth under mechanical constraints.” has not been discussed in the text and no direct data was shown. As it is formulated it sounds that FRET-IBRA has been extensively used in the community rather than on a particular data set. Please reformulate (e.g. specify the lab) or remove.

- Ln 11. “cleanest signal possible” I recommend softening the assertion.

- Ln 145. “without any artifacts” I suggest replacing by something like “without clear artifacts or extreme outliers”

**********

Have the authors made all data and (if applicable) computational code underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data and code underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data and code should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data or code —e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

**********

PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: Yes: Daniel Damineli

Figure Files:

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email us at figures@plos.org.

Data Requirements:

Please note that, as a condition of publication, PLOS' data policy requires that you make available all data used to draw the conclusions outlined in your manuscript. Data must be deposited in an appropriate repository, included within the body of the manuscript, or uploaded as supporting information. This includes all numerical values that were used to generate graphs, histograms etc.. For an example in PLOS Biology see here: http://www.plosbiology.org/article/info%3Adoi%2F10.1371%2Fjournal.pbio.1001908#s5.

Reproducibility:

To enhance the reproducibility of your results, we recommend that you deposit your laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. Additionally, PLOS ONE offers an option to publish peer-reviewed clinical study protocols. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols

References:

Review your reference list to ensure that it is complete and correct. If you have cited papers that have been retracted, please include the rationale for doing so in the manuscript text, or remove these references and replace them with relevant current references. Any changes to the reference list should be mentioned in the rebuttal letter that accompanies your revised manuscript.

If you need to cite a retracted article, indicate the article’s retracted status in the References list and also include a citation and full reference for the retraction notice.

Submitted filename: Report_Details_PCOMPBIOL-D-21-01180R1.pdf

Click here for additional data file.

30 Jan 2022

Submitted filename: Munglani_revision2_response.docx

Click here for additional data file.

16 Feb 2022

Dear Dr. Vogler,

We are pleased to inform you that your manuscript 'Fast and flexible processing of large FRET image stacks using the FRET-IBRA toolkit' has been provisionally accepted for publication in PLOS Computational Biology.

Before your manuscript can be formally accepted you will need to complete some formatting changes, which you will receive in a follow up email. A member of our team will be in touch with a set of requests. Also please address the comments of Reviewer 1 (below) in the final version.

Please note that your manuscript will not be scheduled for publication until you have made the required changes, so a swift response is appreciated.

IMPORTANT: The editorial review process is now complete. PLOS will only permit corrections to spelling, formatting or significant scientific errors from this point onwards. Requests for major changes, or any which affect the scientific understanding of your work, will cause delays to the publication date of your manuscript.

Should you, your institution's press office or the journal office choose to press release your paper, you will automatically be opted out of early publication. We ask that you notify us now if you or your institution is planning to press release the article. All press must be co-ordinated with PLOS.

Thank you again for supporting Open Access publishing; we are looking forward to publishing your work in PLOS Computational Biology.

Best regards,

Dina Schneidman

Software Editor

PLOS Computational Biology

***********************************************************

Reviewer's Responses to Questions

Comments to the Authors:

Please note here if the review is uploaded as an attachment.

Reviewer #1: In this revised version of the manuscript, the authors addressed the main shortcomings raised in the previous revision by applying the pipeline to pollen tube oscillations data and by fixing minor code issues. The present version of the paper and GitHub code offers the cell biology community a significant tool for processing ratiometric data. However, there are still minor points I would like to raise, although I believe they should be implemented at the author’s discretion.

REMAINING REMARKS

Given the significant improvement and number of revised versions the authors have already provided, I emphasize the points below are suggestions that could be implemented if the authors find it in their best interest.

1) Oscillations:

I was hoping the authors would use pollen tube tip oscillations as means to demonstrate the performance of FRET-IBRA to reveal biologically relevant phenomena. As such, ideally, I was expecting to see an additional figure in the paper showing the oscillation extracted with FRET-IBRA preferably in comparison to oscillations extracted with other algorithms. Instead, the authors chose to include part of the result in the GitHub repository, which not only limits the size of the output that can be shared but also does not highlight the performance of FRET-IBRA in terms of revealing clear oscillations. I encourage the authors to show “tip fluorescence vs time” plots instead of the raw output of the pipeline.

2) Limitations of the background subtraction algorithm:

In the previous revision I raised a question about spatial artifacts introduced by the background subtraction algorithm in Brachypodium roots data (specifically “Root_acceptor_back.tif” and “Root_donor_back.tif” found in the GitHub repository). It is clear there are spatial artifacts and this should be discussed, as users have to be aware of this possibility. As such, I did not understand the author’s answer that “Background subtraction, however, works still well, even with very low signals”. I recommend these specific files are inspected carefully and either: a) adequate parameters are used as to not produce spatial artifact; b) they are used as a cautionary example for users (which should be made clear in the text); c) they are excluded altogether.

3) Additional information on the GitHub repository:

I appreciate the authors included README files in GitHub examples, although they should include more explicit information about organism used and source of the data.

**********

Have the authors made all data and (if applicable) computational code underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data and code underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data and code should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data or code —e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

**********

PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: Yes: Daniel Damineli

31 Mar 2022

PCOMPBIOL-D-21-01180R2

Fast and flexible processing of large FRET image stacks using the FRET-IBRA toolkit

Dear Dr Vogler,

I am pleased to inform you that your manuscript has been formally accepted for publication in PLOS Computational Biology. Your manuscript is now with our production department and you will be notified of the publication date in due course.

The corresponding author will soon be receiving a typeset proof for review, to ensure errors have not been introduced during production. Please review the PDF proof of your manuscript carefully, as this is the last chance to correct any errors. Please note that major changes, or those which affect the scientific understanding of the work, will likely cause delays to the publication date of your manuscript.

Soon after your final files are uploaded, unless you have opted out, the early version of your manuscript will be published online. The date of the early version will be your article's publication date. The final article will be published to the same URL, and all versions of the paper will be accessible to readers.

Thank you again for supporting PLOS Computational Biology and open-access publishing. We are looking forward to publishing your work!

With kind regards,

Agnes Pap

PLOS Computational Biology | Carlyle House, Carlyle Road, Cambridge CB4 3DN | United Kingdom ploscompbiol@plos.org | Phone +44 (0) 1223-442824 | ploscompbiol.org | @PLOSCompBiol