TrueProbes: Quantitative Single-Molecule RNA-FISH Probe Design Improves RNA Detection
Curation statements for this article:-
Curated by eLife
eLife Assessment
This useful study introduces a computational pipeline for designing RNA in situ fluorescence hybridization probes that could improve the sensitivity and specificity of RNA detection in cells. While the approach is novel and the preliminary data suggestive, the evidence supporting a clear advantage over existing probe design strategies is incomplete. The work will be of interest to researchers developing or using molecular tools for imaging RNA in cells.
This article has been Reviewed by the following groups
Discuss this preprint
Start a discussion What are Sciety discussions?Listed in
- Evaluated articles (eLife)
Abstract
Abstract
Single-molecule RNA fluorescence in situ hybridization (smRNA-FISH) is a widely used method for visualizing and quantifying RNA molecules in cells and tissues at high spatial resolution. The technique relies on fluorescently labeled oligonucleotide probes that hybridize to target RNA. Accurate quantification depends on high probe specificity to ensure fluorescent signals reflect target RNA binding rather than off-target interactions. Numerous factors, including genome sequence complexity, secondary probe structure, hybridization conditions, and gene expression variability across cell types and lines, influence smRNA-FISH probe efficacy. Existing smRNA-FISH probe design tools have limitations, including narrow heuristics, incomplete off-target assessment, and reliance on “trial-and-error approaches. To address these challenges, we developed TrueProbes, a probe design software platform that integrates genome-wide BLAST-based binding analysis with thermodynamic modeling to generate high-specificity probe sets. TrueProbes ranks and selects probes based on predicted binding affinity, target specificity, and structural constraints. It also incorporates a thermodynamic-kinetic simulation model to provide predictive design metrics and optimize probe performance under user-defined conditions. We benchmarked TrueProbes against several widely used smRNA-FISH design tools and found that it consistently outperformed alternatives across multiple computational metrics and experimental validation assays. Probes designed with TrueProbes showed enhanced target selectivity and superior experimental performance.
Article activity feed
-
eLife Assessment
This useful study introduces a computational pipeline for designing RNA in situ fluorescence hybridization probes that could improve the sensitivity and specificity of RNA detection in cells. While the approach is novel and the preliminary data suggestive, the evidence supporting a clear advantage over existing probe design strategies is incomplete. The work will be of interest to researchers developing or using molecular tools for imaging RNA in cells.
-
Reviewer #1 (Public review):
The authors describe a new computational pipeline designed to identify smFISH probes with improved RNA detection compared to preexisting approaches. smFISH is a powerful and relatively straightforward technique to detect single RNAs in cells at subcellular resolution, which is critical for understanding gene expression regulation at the RNA level. However, existing methods for designing smFISH oligos suffer from several limitations, including off-target binding that produces high background signals, as well as a restricted number of probes that are sufficiently specific to target shorter-than-average mRNAs. To address these challenges, the authors developed TrueProbes, a computational method that aims to minimize off-target-mediated background fluorescence.
Overall, the study addresses a technically relevant …
Reviewer #1 (Public review):
The authors describe a new computational pipeline designed to identify smFISH probes with improved RNA detection compared to preexisting approaches. smFISH is a powerful and relatively straightforward technique to detect single RNAs in cells at subcellular resolution, which is critical for understanding gene expression regulation at the RNA level. However, existing methods for designing smFISH oligos suffer from several limitations, including off-target binding that produces high background signals, as well as a restricted number of probes that are sufficiently specific to target shorter-than-average mRNAs. To address these challenges, the authors developed TrueProbes, a computational method that aims to minimize off-target-mediated background fluorescence.
Overall, the study addresses a technically relevant problem. If improved, this would allow researchers to study gene expression regulation more effectively using single-molecule FISH. However, based on the current presentation of data, it is not yet clear that TrueProbes offers significant advantages over preexisting pipelines. In the following section, I describe some concerns, which should be adequately addressed.
Major Comments:
(1) The manuscript currently presents only one example in which different pipelines were tested to generate probes (targeting ARF4). While the images suggest that both TrueProbes and Stellaris outperform the other pipelines, the comparison is potentially misleading because the number of probes used differs substantially. I recommend that the authors include at least three independent examples in which an equal number of probes are designed across pipelines, so that signal-to-noise can be assessed in a controlled and comparable way. This would allow the probe number to be held constant while directly evaluating performance.
(2) It is also unclear how many biological replicates were performed for the ARF4 experiments. If only a single replicate was included, it is difficult to conclude that TrueProbes consistently outperforms other pipelines in a robust and reproducible manner. I suggest the authors include data from at least three biological replicates with appropriate statistical analysis, and ideally extend this to additional smFISH targets as outlined in Comment 1.
(3) No controls are presented to demonstrate that the TrueProbes-designed smFISH spots are specifically detecting ARF4. The current experiment primarily measures signal-to-noise, but it remains possible that some detected spots do not correspond to ARF4 mRNAs. Since one of the major criteria used by TrueProbes is to limit cross-hybridization, the authors should perform ARF4 knockdown experiments and demonstrate that nearly all ARF4 smFISH signal is lost. A similar approach should be applied to the additional examples recommended in Comment 1.
(4) In the limitations of the study, the authors note that "RNA secondary and tertiary structures are not included, which may lead to inaccuracies if binding sites are structurally occluded." However, I am not convinced that this is a true limitation, since formamide in the smFISH protocol should denature secondary structures and allow oligo access to the RNA. I recommend that the authors comment on this point and clarify whether secondary structure poses a practical limitation in smFISH probe design.
(5) The authors also correctly acknowledge in their limitations that "RNA-protein interactions, which can modulate accessibility of the transcript, are not modeled." I suggest referencing relevant studies on this issue, particularly Buxbaum et al. (2014, Science), which would provide important context.
-
Reviewer #2 (Public review):
Summary:
Hughes et al present a new single-molecule RNA fluorescence in situ hybridization (smFISH) probe design software, termed "TrueProbes" in this manuscript. They claim that all existing smFISH (and variants) probe design software packages have limitations that ultimately impact experimental performance. The author's claim to address the majority of these limitations in TrueProbes by introducing multiple computational steps to ensure high-quality probe design. The manuscript's goal is clear, and the authors provide some evidence by designing and targeting one gene. Overall, the manuscript lacks rigorous evidence to support the claims, does not demonstrate its suitability for a variety of smFISH-type experiments, and some of the provided quantification data are unclear. While TrueProbes clearly has …
Reviewer #2 (Public review):
Summary:
Hughes et al present a new single-molecule RNA fluorescence in situ hybridization (smFISH) probe design software, termed "TrueProbes" in this manuscript. They claim that all existing smFISH (and variants) probe design software packages have limitations that ultimately impact experimental performance. The author's claim to address the majority of these limitations in TrueProbes by introducing multiple computational steps to ensure high-quality probe design. The manuscript's goal is clear, and the authors provide some evidence by designing and targeting one gene. Overall, the manuscript lacks rigorous evidence to support the claims, does not demonstrate its suitability for a variety of smFISH-type experiments, and some of the provided quantification data are unclear. While TrueProbes clearly has potential, more data is required, or the authors should tone down the claims.
Strengths:
(1) The problem is well-articulated in the abstract and the introduction.
(2) Figures 3 and 4 follow a consistent color scheme where each probe design method has its own color, which helps the reader visually compare methods.
(3) The authors compared multiple probe design software packages both computationally and experimentally.
(4) TrueProbes does produce visually and quantitatively better results when compared to 2 of the 4 existing smFISH probe design packages (Paintshop and MERFISH panel designer).
(5) The authors introduce a comprehensive steady-state thermodynamic model to help optimally guide probe design.
Weaknesses:
(1) The abstract describes the problem well and introduces the solution (the TrueProbes software), but fails to provide specific ways in which the TrueProbes software performs better. The authors state that "...[TrueProbes] consistently outperformed alternatives across multiple computational metrics and experimental validation assays", but specific, quantitative evidence of improved performance would strengthen the statement.
(2) The text claims that TrueProbes outperforms all other probe design software, but Figure 3 indicates that TrueProbes has neither the greatest number of on-target binding nor the lowest number of off-target binding. The data in Figure 3 does not support the claims made in the text. Specifically, the authors claim that "RNA FISH Experimental Results Demonstrate that Off Target and Binding Affinity Inclusive Probe Design Improve RNA FISH Signal Discrimination" (lines 217-218). However, despite their claim that Stellaris and Oligostan-HT produce more off-target probes when evaluated with the TrueProbes framework, the experiment results are nearly identical. The authors should consider modifying their claims or performing new experiments that more clearly demonstrate their claims.
(3) The bar graphs in Figure 3 do not seem to agree with the probability graphs in Figure 4. For example, Figure 3 indicates that Stellaris probes have higher off-target binding than TrueProbes; however, in Figure 4, their probability graphs lie almost on top of each other.
(4) The authors performed validation for only one gene (ARF4), because "...it had the highest gene expression (in TPM units) and the fewest isoforms among all candidate genes for the Jurkat cell line" (lines 176-177). While the results do look good, this is a minimal use case and does not really showcase the power of their method. One experiment that could be helpful would be two-color (or more) smFISH in tissue, where the chances for off-target binding contributing to higher errors are much greater than in an adherent cell line.
(5) A common strategy for both smFISH and highly multiplexed methods is to use secondary DNA oligos with dye molecules instead of direct conjugation. Given that this is a primary design goal of PaintSHOP and the Zhuang lab's MERFISH probe design code, it would be helpful to demonstrate that TrueProbes can design a two-layer probe strategy for high-quality RNA-FISH labeling.
(6) The authors claim, "For every probe set, TrueProbes can simulate expected smRNA FISH outcomes including optimal probe, RNA, and salt concentrations and optionally account for probe secondary structure, hybridization temperature, multiple targets, fluorophore choice, DNA, nascent RNA, and photon count statistics (Figures S2A, S2B). The model can be used to generate predictions for temperature and cell line sensitivity, multi-target discrimination, multiple fluorophore colocalization; when provided transcript expression levels and probe/background intensity, it can start to generate predictions for spot intensity, background, signal to noise ratio, and false negative rates (Figure S2C)." (lines 156-163). Figure S2 is a flow chart and does not provide evidence for any of these items. The authors should provide evidence for these claims, either as a figure or an example script in their software repository. If that is not possible, then it should be removed.
(7) All thermodynamic equations are performed at steady state. The authors do not justify this assumption, and there is no discussion of the potential impacts of either low molecule numbers or violations of the well-mixed assumption. Can the authors please include a discussion on the potential impacts non non-steady state dynamics?
-
Reviewer #3 (Public review):
Summary:
This manuscript introduces a new platform termed "TrueProbes" for designing mRNA FISH probes. In comparison to existing design strategies, the authors incorporate a comprehensive thermodynamic and kinetic model to account for probe states that may contribute to nonspecific background. The authors validate their design pipeline using Jurkat cells and provide evidence of improved probe performance.
Strengths:
A notable strength of TrueProbes is the consideration of genome-wide binding affinities, which aims to minimize off-target signals. The work will be of interest to researchers employing mRNA FISH in certain human cell lines.
Weaknesses:
However, in my view, the experimental validation is not sufficient to justify the broad claims of the platform. Given the number of assumptions in the model, …
Reviewer #3 (Public review):
Summary:
This manuscript introduces a new platform termed "TrueProbes" for designing mRNA FISH probes. In comparison to existing design strategies, the authors incorporate a comprehensive thermodynamic and kinetic model to account for probe states that may contribute to nonspecific background. The authors validate their design pipeline using Jurkat cells and provide evidence of improved probe performance.
Strengths:
A notable strength of TrueProbes is the consideration of genome-wide binding affinities, which aims to minimize off-target signals. The work will be of interest to researchers employing mRNA FISH in certain human cell lines.
Weaknesses:
However, in my view, the experimental validation is not sufficient to justify the broad claims of the platform. Given the number of assumptions in the model, additional experimental comparisons across probe design methods, ideally targeting transcripts with different expression levels, would be necessary to establish the general superiority of this approach.
-
Author response:
Reviewer #1 (Public Review):
The authors describe a new computational pipeline designed to identify smFISH probes with improved RNA detection compared to preexisting approaches. smFISH is a powerful and relatively straightforward technique to detect single RNAs in cells at subcellular resolution, which is critical for understanding gene expression regulation at the RNA level. However, existing methods for designing smFISH oligos suffer from several limitations, including off-target binding that produces high background signals, as well as a restricted number of probes that are sufficiently specific to target shorter-than-average mRNAs. To address these challenges, the authors developed TrueProbes, a computational method that aims to minimize off-target-mediated background fluorescence.
Overall, the study addresses a …
Author response:
Reviewer #1 (Public Review):
The authors describe a new computational pipeline designed to identify smFISH probes with improved RNA detection compared to preexisting approaches. smFISH is a powerful and relatively straightforward technique to detect single RNAs in cells at subcellular resolution, which is critical for understanding gene expression regulation at the RNA level. However, existing methods for designing smFISH oligos suffer from several limitations, including off-target binding that produces high background signals, as well as a restricted number of probes that are sufficiently specific to target shorter-than-average mRNAs. To address these challenges, the authors developed TrueProbes, a computational method that aims to minimize off-target-mediated background fluorescence.
Overall, the study addresses a technically relevant problem. If improved, this would allow researchers to study gene expression regulation more effectively using single-molecule FISH. However, based on the current presentation of data, it is not yet clear that TrueProbes offers significant advantages over preexisting pipelines. In the following section, I describe some concerns, which should be adequately addressed.
Major Comments:
(1) The manuscript currently presents only one example in which different pipelines were tested to generate probes (targeting ARF4). While the images suggest that both TrueProbes and Stellaris outperform the other pipelines, the comparison is potentially misleading because the number of probes used differs substantially. I recommend that the authors include at least three independent examples in which an equal number of probes are designed across pipelines, so that signal-to-noise can be assessed in a controlled and comparable way. This would allow the probe number to be held constant while directly evaluating performance.
This is an important observation. We have already addressed this issue in Figures 3E-G and Supplementary Figure 4E-G, where we plotted the number of OFF-targets for each ON-target probe. If we select longer genes to ensure an equal number of designed probes with strong signals, we will still end up with the same number of ON-target probes. Consequently, Figures 3B-D and 3E-G would show similar trends, albeit with different values on the y-axis. Additionally, we will conduct an analysis using Stellaris at its highest probe design stringency setting to compare the software under its strictest design conditions. Additional experiments are outside the scope of the current manuscript.
(2) It is also unclear how many biological replicates were performed for the ARF4 experiments. If only a single replicate was included, it is difficult to conclude that TrueProbes consistently outperforms other pipelines in a robust and reproducible manner. I suggest the authors include data from at least three biological replicates with appropriate statistical analysis, and ideally extend this to additional smFISH targets as outlined in Comment 1.
Three biological replicates were utilized for the ARF4 experiments. As stated in the original submission, the average data from all three replicates is presented in Figure 4, while the data for each individual replicate can be found in Figure S5. Statistical analyses were conducted for both the pooled data in Figure 4 and the individual data in Figure S5. The results of all statistical calculations are detailed in Supplemental Table 1. We will update the text to clearly indicate the number of biological replicates and the outcomes of the statistical analysis.
(3) No controls are presented to demonstrate that the TrueProbes-designed smFISH spots are specifically detecting ARF4. The current experiment primarily measures signal-to-noise, but it remains possible that some detected spots do not correspond to ARF4 mRNAs. Since one of the major criteria used by TrueProbes is to limit cross-hybridization, the authors should perform ARF4 knockdown experiments and demonstrate that nearly all ARF4 smFISH signal is lost. A similar approach should be applied to the additional examples recommended in Comment 1.
Thank you for your suggestion. Currently, we lack the expertise in our lab to conduct such experiments, so they are beyond the scope of this manuscript. However, we will create additional supplementary figures to demonstrate that the likelihood of false positives is low, based on the assumption that current publicly available BLAST algorithms, genome annotations, and reference transcription expression data are accurate.
We will include a comparison in our supplementary materials showing the off-target RNA that can bind the highest number of probes simultaneously for each software. Additionally, we will perform a correlation analysis to illustrate the relationship between spot intensity for different software and the number of probes they design. This will help us estimate how the number of probes bound to RNA correlates with expected spot intensity ranges.
Using this information, along with autofluorescence background intensity measurements from no-probe controls, we will estimate the minimum number of probes that need to bind to targets to be detected as single spots. If this minimum is higher than the maximum number of simultaneous off-target probe bindings, we anticipate that the detected spot signal will primarily reflect ARF4 rather than other transcripts.
(4) In the limitations of the study, the authors note that "RNA secondary and tertiary structures are not included, which may lead to inaccuracies if binding sites are structurally occluded." However, I am not convinced that this is a true limitation, since formamide in the smFISH protocol should denature secondary structures and allow oligo access to the RNA. I recommend that the authors comment on this point and clarify whether secondary structure poses a practical limitation in smFISH probe design.
Thank you for pointing this out. We will revise the manuscript to clarify: "We did not include RNA secondary and tertiary structures in the model because the use of formamide in RNA-FISH experiments denatures these structures, allowing oligonucleotides to access the RNA."
(5) The authors also correctly acknowledge in their limitations that "RNA-protein interactions, which can modulate accessibility of the transcript, are not modeled." I suggest referencing relevant studies on this issue, particularly Buxbaum et al. (2014, Science), which would provide important context.
Thank you for highlighting the literature that supports this limitation. We will include Buxbaum et al. (2014, Science) and additional studies that discuss how RNA-protein interactions can affect RNA-FISH experiments.
Reviewer #2 (Public review):
Summary:
Hughes et al present a new single-molecule RNA fluorescence in situ hybridization (smFISH) probe design software, termed "TrueProbes" in this manuscript. They claim that all existing smFISH (and variants) probe design software packages have limitations that ultimately impact experimental performance. The author's claim to address the majority of these limitations in TrueProbes by introducing multiple computational steps to ensure high-quality probe design. The manuscript's goal is clear, and the authors provide some evidence by designing and targeting one gene. Overall, the manuscript lacks rigorous evidence to support the claims, does not demonstrate its suitability for a variety of smFISH-type experiments, and some of the provided quantification data are unclear. While TrueProbes clearly has potential, more data is required, or the authors should tone down the claims.
We appreciate the reviewer’s thoughtful feedback. We will revise the text to ensure that all claims are backed by computational or experimental evidence. For claims that do not have supporting results, we will relocate them to the discussion section as potential future extensions. Since our probe design is open access, both we and the community can further develop our codes as needed.
Strengths:
(1) The problem is well-articulated in the abstract and the introduction.
(2) Figures 3 and 4 follow a consistent color scheme where each probe design method has its own color, which helps the reader visually compare methods.
(3) The authors compared multiple probe design software packages both computationally and experimentally.
(4) TrueProbes does produce visually and quantitatively better results when compared to 2 of the 4 existing smFISH probe design packages (Paintshop and MERFISH panel designer).
(5) The authors introduce a comprehensive steady-state thermodynamic model to help optimally guide probe design.
We like to thank the reviewer for pointing out the strength of the manuscript.
Weaknesses:
(1) The abstract describes the problem well and introduces the solution (the TrueProbes software), but fails to provide specific ways in which the TrueProbes software performs better. The authors state that "...[TrueProbes] consistently outperformed alternatives across multiple computational metrics and experimental validation assays", but specific, quantitative evidence of improved performance would strengthen the statement.
Thank you for acknowledging the clarity of the abstract and introduction. We will revise the abstract to provide more specific details on how TrueProbes outperforms other software. Additionally, we will include specific computational and experimental metrics that demonstrate TrueProbes' improved performance compared to other software.
(2) The text claims that TrueProbes outperforms all other probe design software, but Figure 3 indicates that TrueProbes has neither the greatest number of on-target binding nor the lowest number of off-target binding. The data in Figure 3 does not support the claims made in the text. Specifically, the authors claim that "RNA FISH Experimental Results Demonstrate that Off Target and Binding Affinity Inclusive Probe Design Improve RNA FISH Signal Discrimination" (lines 217-218). However, despite their claim that Stellaris and Oligostan-HT produce more off-target probes when evaluated with the TrueProbes framework, the experiment results are nearly identical. The authors should consider modifying their claims or performing new experiments that more clearly demonstrate their claims.
In Figure 3, we aim to convey two main points.
The first point is to compare the number of ON-target probes designed by each software using their most stringent design criteria (Figure 3A). Currently, we are using a medium strict design criterion for Stellaris (level 3). As shown in the new supplementary figure XX, when we apply the most stringent design criteria for Stellaris (level 5), the number of ON-target probes decreases to XX probes. This clearly indicates that, based on theoretical calculations, TrueProbes can design more probes than any of its competitors.
The second point is to compare the number of OFF-targets produced by each probe design. To illustrate this, we used two different metrics. In Figures 3B-D, we compare the total number of probes bound to OFF-target RNA. However, since each software generates a different number of ON-target probes, the number of OFF-targets may vary simply due to the differences in ON-target probe counts. Therefore, we introduced a second metric to compare OFF-targets. In Figures 3E-G, we present the number of OFF-targets normalized by the number of ON-targets. Using this metric, TrueProbes shows the lowest number of OFF-targets. We will updat the manuscript to clarify this point.
Regarding the experiments and their comparison to theoretical calculations: The theoretical calculations consider only the reference DNA and RNA genomes along with the oligonucleotide sequences for the probes. We then use a thermodynamic model to identify ON- and OFF-targets. Thus, these theoretical calculations represent an upper bound on the maximum possible number of ON-targets and the minimum number of OFF-targets. All other design software evaluated in this manuscript relies on the same or less reference data and makes certain assumptions. None of these methods quantitatively compare their computational designs with experimental results; they simply design probes based on unverified assumptions, conduct experiments, and present spot data to conclude that their probe designs are effective.
We will update the manuscript to clarify the goals of the theoretical model and its relationship to the experiments. Future work will be necessary to enhance our theoretical model to fully account for additional aspects of RNA-FISH experiments (e.g., formaldehyde crosslinking, hybridization conditions, washing steps) to better predict the experimental data shown in Figure 4. We will also adjuste our claims to accurately reflect the current capabilities of our theoretical framework and its relation to experimental outcomes.
(3) The bar graphs in Figure 3 do not seem to agree with the probability graphs in Figure 4. For example, Figure 3 indicates that Stellaris probes have higher off-target binding than TrueProbes; however, in Figure 4, their probability graphs lie almost on top of each other.
The predictions in Figure 3 regarding the number of probe off-target binding events, based on reference gene expression data, do not necessarily encompass all the information required to predict RNA-FISH signal intensity. Therefore, these predictions should not be expected to translate directly into the experimental results shown in Figure 4, particularly concerning the background signal.
While our software aims to minimize off-target probe binding, this does not automatically lead to a reduction in off-target background signal. Numerous other factors influence the spot background and overall signal-to-noise ratio (SNR) performance, beyond just probe-target binding interactions. Although we strive to minimize off-target background through probe binding, this approach is not designed to directly predict the SNR. Extending the computational analysis of probe binding dynamics to RNA-FISH signal intensity dynamics is beyond the scope of this study.
We have revised our text to clearly separate computational results from experimental results into two distinct sections. We will use different terminology to describe the outcomes of computational performance versus experimental performance, reducing potential confusion between these two aspects. Additionally, we will clarify our conceptual overview in Figure 1 regarding traditional probe design limitations related to sensitivity and specificity. We will specify how the signal from the number of probes bound to ON-target RNA, relative to those bound to OFF-targets and cellular autofluorescence, translates—either linearly or non-linearly—into the signal-to-noise ratio.
(4) The authors performed validation for only one gene (ARF4), because "...it had the highest gene expression (in TPM units) and the fewest isoforms among all candidate genes for the Jurkat cell line" (lines 176-177). While the results do look good, this is a minimal use case and does not really showcase the power of their method. One experiment that could be helpful would be two-color (or more) smFISH in tissue, where the chances for off-target binding contributing to higher errors are much greater than in an adherent cell line.
Thank you for highlighting these valuable experiments. Currently, our lab lacks the expertise to generate tissue samples beyond culturing cells. Additionally, implementing a two-color probe design in tissues containing different cell types with unknown expression levels presents further challenges. Due to these limitations, designing and conducting two-color experiments in tissue samples is beyond the scope of the current manuscript, but we plan to pursue this in the future.
(5) A common strategy for both smFISH and highly multiplexed methods is to use secondary DNA oligos with dye molecules instead of direct conjugation. Given that this is a primary design goal of PaintSHOP and the Zhuang lab's MERFISH probe design code, it would be helpful to demonstrate that TrueProbes can design a two-layer probe strategy for high-quality RNA-FISH labeling.
Thank you for bringing this to our attention. TrueProbes is currently designed and tested specifically for primary smRNA-FISH probes. Our focus is on demonstrating a new approach to designing these probes without the added complexities of secondary probes and multiplexing. Future work will expand on this foundation to incorporate secondary probe detection and transcript multiplexing.
(6) The authors claim, "For every probe set, TrueProbes can simulate expected smRNA FISH outcomes including optimal probe, RNA, and salt concentrations and optionally account for probe secondary structure, hybridization temperature, multiple targets, fluorophore choice, DNA, nascent RNA, and photon count statistics (Figures S2A, S2B). The model can be used to generate predictions for temperature and cell line sensitivity, multi-target discrimination, multiple fluorophore colocalization; when provided transcript expression levels and probe/background intensity, it can start to generate predictions for spot intensity, background, signal to noise ratio, and false negative rates (Figure S2C)." (lines 156-163). Figure S2 is a flow chart and does not provide evidence for any of these items. The authors should provide evidence for these claims, either as a figure or an example script in their software repository. If that is not possible, then it should be removed.
The supplemental information of the article will be updated to include figures that illustrate predictions for each capability currently offered by TrueProbes, along with the scripts used to generate these predictions. Any capabilities that do not have corresponding scripts will be removed from this section and instead referred to as potential improvements or future additions to the TrueProbes framework in the discussion section.
(7) All thermodynamic equations are performed at steady state. The authors do not justify this assumption, and there is no discussion of the potential impacts of either low molecule numbers or violations of the well-mixed assumption. Can the authors please include a discussion on the potential impacts non non-steady state dynamics?
Thermodynamic equations are calculated at steady state because RNA-FISH hybridization reactions typically last from eight to twenty hours. This duration allows probes adequate time to localize to their targets and reach binding equilibrium, based on current estimates of DNA oligonucleotide association and dissociation rate constants. We will address the potential violation of the well-mixed assumption in the assumptions and limitations section, specifically discussing how RNA localization can affect the spatial distribution of both on-target and off-target probes within cells, which may disrupt the well-mixed condition.
Low molecule numbers are not a significant concern, as probe DNA oligonucleotide concentrations in RNA-FISH protocols are much higher than the number of transcripts present in cells, by several orders of magnitude.
The assumptions and limitations section will be revised to clearly state: “Probe hybridization reactions were computed at steady state because most RNA-FISH protocols utilize probe hybridization incubation steps lasting over eight hours, which should provide sufficient time to reach equilibrium based on current estimates of forward and reverse reaction rate constants. Predictions from the equilibrium model may be less accurate for RNA-FISH experiments with shorter hybridization times, where non-steady state dynamics can result in different transient outcomes depending on the duration of hybridization.”
Reviewer #3 (Public review):
Summary:
This manuscript introduces a new platform termed "TrueProbes" for designing mRNA FISH probes. In comparison to existing design strategies, the authors incorporate a comprehensive thermodynamic and kinetic model to account for probe states that may contribute to nonspecific background. The authors validate their design pipeline using Jurkat cells and provide evidence of improved probe performance.
Strengths:
A notable strength of TrueProbes is the consideration of genome-wide binding affinities, which aims to minimize off-target signals. The work will be of interest to researchers employing mRNA FISH in certain human cell lines.
Weaknesses:
However, in my view, the experimental validation is not sufficient to justify the broad claims of the platform. Given the number of assumptions in the model, additional experimental comparisons across probe design methods, ideally targeting transcripts with different expression levels, would be necessary to establish the general superiority of this approach.
We will revise our text to make our claims more specific and clearer, avoiding overgeneralizations and ensuring that all claims are adequately supported by the data we present.
-
-
-
