Multiplexed detection of viral antigen and RNA using nanopore sensing and encoded molecular probes

We report on single-molecule nanopore sensing combined with position-encoded DNA molecular probes, with chemistry tuned to simultaneously identify various antigen proteins and multiple RNA gene fragments of SARS-CoV-2 with high sensitivity and selectivity. We show that this sensing strategy can directly detect spike (S) and nucleocapsid (N) proteins in unprocessed human saliva. Moreover, our approach enables the identification of RNA fragments from patient samples using nasal/throat swabs, enabling the identification of critical mutations such as D614G, G446S, or Y144del among viral variants. In particular, it can detect and discriminate between SARS-CoV-2 lineages of wild-type B.1.1.7 (Alpha), B.1.617.2 (Delta), and B.1.1.539 (Omicron) within a single measurement without the need for nucleic acid sequencing. The sensing strategy of the molecular probes is easily adaptable to other viral targets and diseases and can be expanded depending on the application required.

How reproducible are the results across different nanopores?Is the data in eacy figure from single pores or aggregated across many?It is not clear whether the blockage levels, for example, are the same in different pores.Figures S5 and S6 have very similar DNA blockages, but S13 shows larger blockages -is that the pore?Are the triplicate-result binding curves all taken in the same pore?If not, are the pores nearly identical in conductance?Throughout the paper, it would be very useful to know whether these experiments are all being performed in just a few, or many, pores, as this is one of the major challenges in the nanopore field.Do the results from clinical samples (for the RNA, at least) correspond to concentrations that match qRT-PCR results?It may be too late to collect these, but the data is very close to being able to verify the ability to determine viral load.This could be an important and exciting additional result to include.
In a similar vein, relatively little data from the clinical samples is shown in terms of the events and scatterplots characterizing the nanopore data.The pre-clinical samples are described and a bit is shown, but the actual clinical samples only have reported statistics.Showing more of the raw data and events in SI figures would be very informative and interesting to the reader.
The authors are surely aware that many nanopore papers draw questions about the ease of use and requirements for pre-nanopore PCR.I suggest that they carefully re-consider some of the wording in the intro regarding the need for a new diagnostic approach without PCR, fast and on-site, simple, etc, because here they are still using PCR, and it is a highly complex detection scheme.
What is the source of the pseudovirus?Methods says just "suppliers".

FIGURES
Figures are generally very nicely laid out and designed.The extensive and thorough SI figures are particularly appreciated, as is the consistent color coding for different assays / samples that is maintained across all figures.A few minor comments and suggestions follow.
Figure 2: what are the protein concentrations for each sample in 2e? Figure 3 requires scale bars for time and current blockage for panel a.Should show an additional panel similar to 1c.Also, in d, the binning in time is too large and does not show the shape of the distribution.Fig. 2 and 3: what is the source of the S and N proteins used here?Are they from the pseudovirus?How are they purified and verified?Across all figures: it is good practice to include the number of events either on the figure or in the caption for histograms and scatter plots.Figures S8 and S9 have this but none of the others seem to.
Reviewer #3 (Remarks to the Author): This manuscript by Ren et al. reports multiplex detection of SARS-CoV-2 antigens and RNAs using glass pipette-based solid-state nanopores and dsDNA probes.If successfully demonstrated, the potential clinical significance of this technology is clear, and the technology can be translational.However, I am afraid the current manuscript is not suitable for publication due to several major flaws in the experimental design and the data analysis as listed below.
Major Concerns: In the Introduction section, the authors failed to explain the novelty of this work in comparison to similar previous papers (JACS https://doi.org/10.1021/jacs.2c13465;Nature Communications (2021) 12:3515; Nature Nanotechnology (2023) 18:290-298).The first 2 papers are from the same group of authors.It is understandable that the first and the third papers were not cited as they are new, but the second paper was only cited without any detailed explanation.Please clearly explain the novelty of this work.
About DNA folding: There is partial folding of dsDNA probes, why not optimize the probe sequence or use smaller nanopores to try to avoid folding?
Please double check "However, these secondary peaks, due to carrier folding, could be readily identified by their significantly smaller blockade current of secondary amplitude (90.0 ± 8.6 pA)…" Which structure causes secondary amplitude (90.0 ± 8.6 pA), S protein or folded DNA?
Author claimed to demonstrate "smaller cross-section of folded DNA double-helix compared to the size of the S protein" in Figure S7a, but the figure only shows 3D structure of S protein without dimension.It is the authors responsibility to demonstrate a clear comparison of dimensions of folded DNA vs. S protein.
Related to the previous comments, even with a comparison of dimensions, it is still hardly a direct evidence that the partially elevated peak is due to partially folded DNA probe.An experimental comparison between signals from a known folded dsDNA and a known unfolded dsDNA may help.

About multiplex gene detection:
Figure 4e shows only one calibration curve.Is this curve used for all 3 genes?It seems, in Figure 4d, that responses to different genes at the same concentration are different.In addition, how are the SD (error bars) calculated if 3 genes are sharing the same curve?
Author mentioned that "For all three genes, 35-cycle PCR amplification was performed…" After reviewing the Method section, it seems RT-PCR was done to amplify the target genes in clinical samples before nanopore measurement.In this case, the nanopore experiments seems redundant, as these target genes could be directly quantified by RT-qPCR, which combines amplification and quantitative detection into one step.Even though the author argues that their method has "approximately two orders of magnitude lower LOD than gold standard RT-qPCR", clinical samples tested were all confirmed RT-qPCR positives.Without testing patients with false negative RT-qPCR, the benefit of adding many complex steps for nanopore measurement is not clear.

About preclinical test in saliva:
Author stated that "S and N protein was spiked in the saliva (with a final concentration of 20 nM)".How is the concentration calculated? in what solution/matrix?Author stated that "According to the calibration curve in Fig. 2d, we estimated the concentration of S protein to be approximately 0.5 pM, Supplementary Fig. S31."From the text related to Figure 2d, the calibration curve was not established using the same solution used in the Pseudovirus experiment, which includes "lysis and extraction buffer" etc.The calibration curve in Figure 2d cannot be used here.In addition, author only calculated LOD as 3σ above the background but did not calculate LOQ (limit of quantification).If LOQ is greater than 0.5pM, then the assay cannot quantify at 0.5pM.Again, Figure 6b results are based on a calibration curve established in a less complex solution (Figure 4e).It appears that Figure 6av is the same as Figure 2d and 3b; Figure 6bv is the same as Figure 4e.The same data should not be presented twice with different figure number references.This is not acceptable.

About clinical test:
The ethics statement is inadequate.
Details about how each patient sample was PCR confirmed are needed.
Again, clinical samples are tested in a different buffer/matrix environment.Fortunately, there is no concentration calculation in this section, only binding ratio was presented.
It is premature to conclude that the test can detect mutations reliably with only 5 clinical samples per mutation.The authors should make this clear.
In the Conclusion section, although the multiplex detection capacity is plausible, current results do not support the claim of "improved accuracy and ultra-high sensitivity".There is no comparison between this assay and benchmark methods, so improved accuracy cannot be claimed.Neither analytical sensitivity nor clinical sensitivity can be rigorously evaluated in this study.Analytical sensitivity should be assessed by calibration curves established using the same buffer as in real-world applications.Clinical sensitivity needs to be assessed using more samples and, more importantly, in parallel with clinical test results and patient information that can help reliably analyze the results (e.g.sample collection time, treatment time, symptoms, etc.)

Minor Comments:
The author claimed that "The binding ratio for S protein increased initially and plateaued after 20 nM."However, in Figure 2d, the binding ratio of 20 nM and 200 nM showed increase.Figure S12b should be N protein?According to Figure 4b(i), in Figure 4c(i) the N site is positive, but the legend states negative.Similar inconsistency is also seen in Figure 4c(ii).
What is the Z axis title and unit in Figure 4d? Figure 5e is confusing.I suggest matching the target of each site (colored blocks on the top) to its fractional position.For example, N at last etc.
Figure S30 indicates ~9% events are short dwell time events suspected to be induced by background molecules.However, in the first 200 events acquired from saliva indicated in Figure S29, none of them were from background molecules.This is highly unlikely to happen statistically.
"We also demonstrated sensitivity down to 1 copy of RNA per 100 μl via a preamplified step in nasal/throat swab patient samples."sounds confusing.It suggests 1 copy/100 μl sensitivity without the PCR amplification which is not the case.
Please double check all references.Reference numbers are not appearing in the correct order in the text.

REVIEWER Response
Reviewer #1 (Remarks to the Author): The manuscript entitled "Multiplexed detection of viral antigen and RNA using nanopore sensing and encoded molecular probes" by Edel and co-authors presented a single-molecule multiplexed detection of viral antigen and RNA using nanopore sensing and encoded molecular probes.Taking the advantage of nanopore resolution, they can encode the probe in different position on the DNA strand and identify multiple target simultaneously.The sensitivity and selectivity reported is remarkable, in particular, they demonstrated that target RNA fragment with single-base mutation can be distinguished, which allows for the identification of different variants of SARS-CoV-2, including Delta and Omicron.Interestingly, the results from the clinical patient samples show great potential in the practical uses for this method.Overall, this is a well written report with an interesting idea.The supporting information are useful.The cited literature seems appropriate.The quality of figures is good.Therefore, I would like to recommend the publication of this manuscript after a minor revision.
1.This manuscript involved sophisticated data analysis, it would be better for the authors to illustrate in a more detailed way to demonstrate how the data has been extracted and analysed.
We have added a section in supporting information about data analysis explaining the analysis procedure.
We have revised the manuscript.All 'wild type' has been changed to 'wild-type' 3. Since the aptamer sequence is new, the authors are recommended to provide the detailed SELEX process in Method section or Supplementary Information.
Thanks for the good suggestion.We have included a detailed description of the SELEX process in the Supplementary Information.
4. In Supplementary Figure 2c, there's 8 T linker, however, in the caption, it is 6 T. Please double check and clarify this.Reviewer #2 (Remarks to the Author): Ren and co-workers report an exciting application of nanopore sensors for amplification-free detection of SARS-CoV-2 proteins (S and N) and transcribed RNA for the S, N, and ORF1b genomic locations.To detect proteins, the authors conjugate long dsDNA to aptamers that specifically bind the protein of interest, and detect it as a secondary peak on the translocation blockage.The peak location within the event reports on the type of construct passing through, so that constructs with aptamers positioned at different locations (center, end, off-center) may be used to multiplex detection.To detect RNA from different gene transcripts, they create secondary peaks using streptavidin bound to a biotinylated oligo complementary to a specific RNA sequence, while a dangling oligo on the larger DNA construct is complementary to a nearby sequence on that same RNA.Using multiple oligo positions on one construct allows multiplexing.They further show discrimination among different mutant variations using oligos optimized for each task.After checking that their assay detects a pseudovirus as well as synthetic viral RNA, they also test patient samples for wild-type, Delta, and Omicron variants.
On the whole this work is novel, is careful, and is thorough.It is not the first nanopore COVIDdetection paper but they do have a great approach that seems robust.I do have some suggestions and questions for the authors, mostly related to choices in data acquisition / analysis / interpretation, but once these are addressed I recommend that this work be published promptly.
Can the authors comment on / explain why their streptavidin tags show such a narrow distribution of fractional peak position?This is surprising given the broad distribution for the N protein and broad distributions observed in other protein-DNA complex nanopore measurements.Is the fractional distribution of the S protein constructs also broad (or has a long tail)?Also, some discussion of how the observed secondary peak amplitudes correspond to the sizes and chemical properties of the various proteins attached to the DNA would be nice.Small point: you may want to mention that it is an S-protein trimer.
For N protein and S protein detection, as it is straightforward to differentiate the sub-peak in the middle or in the end, we took all the translocation events (including folding and unfolding) into account to ensure the accuracy for binding ratio.As the folding of DNA carrier would shift the fractional position within the whole event (as shown in the example below); therefore, the distribution for N protein fractional position is broader.However, for the streptavidin binding experiment, we need to differentiate the 0.2 and 0.5 point, which is used to measure the mutants.We discarded the folding events for the statistics as it would affect the accuracy of determining the fractional positions of a given target.Therefore, the fractional peak position shows narrower as it should be.
We have addressed the comparison of secondary peak amplitudes for the S and N proteins in the main text on Page 7: "These secondary peaks exhibit an average amplitude of 85.2 ± 23.8 pA and a dwell time of 25 ± 18 μs.It is worth mentioning that a fraction of the observed signal events corresponds to partial folding, both in the absence and presence of S protein (Supplementary Fig. S4).However, these secondary peaks resulting from probe folding can be easily distinguished by their significantly smaller blockade current of 32.7 ± 9.9 pA, as shown in Fig. 2c and Supplementary Fig. S5-6.This smaller current can be attributed to the smaller width of the folded DNA double-helix (compared to the size of the S protein, (see Supplementary Fig. S7a for the dimensions of S protein).".
We are using the S1 domain of S protein, not S-protein trimer, we have now clarified this in the main text.
The authors cite a fairly large range of nanopore sizes (15nm +/-3 nm), whose area should vary by more than a factor of 2. How is this range reconciled with the very small range of conductances observed?Were only pores with certain conductances selected for experiments?
The nanopore size range is calculated from SEM images however we use nanopore conductance measurements and peak current SNR for the carrier molecule as indicator of nanopore device-todevice variation.As such we try to use nanopores with minimal conductance deviation (less than 4.6%) in the same set of experiments (for example for S protein detection) to ensure the signals are comparable, see Supplementary Fig. 3; but it should be noted that the pore size could also be different.We have added one section in supporting information about using IV to calculate the pore size.
It is currently not completely clear how prominent the peak for the N protein is relative to the translocation.From the example events in Figure 3, one is led to believe that these peaks are prominent.However, From S17, it appears that the N peak is below a substantial fraction of the regular (folded?)DNA events.Note that S17 shows the DNA mean at 0.09 nA with a reasonably broad distribution and the N protein peak at 0.1 nA with a very narrow distribution.Similarly, in S30 the N protein clusters well within the regular DNA translocation cluster.This point is related to the questions and suggestions regarding transparency on the peak-finding algorithm mentioned below.
The width or height of the subpeaks does not impact the conclusions drawn from the data because we rely on the ratio of bound to unbound events to distinguish positive and negative events.For the height of N protein signal, it is within the DNA translocation cluster because the size of N protein is small.In Figure S17, since the number of detected N protein events (n = 60) is significantly lower than that of the unbound DNA carrier (n = 1877), the distribution of N protein numbers appears narrower compared to that of the DNA carrier (folded and unfolded).This pattern is also observed in Figure S30.This circumstance actually motivated us to design the N protein DNA carrier with the binding site in the middle instead of at the end, as it would have been challenging to differentiate the protein binding signal from the DNA folding signal.We acknowledge the need to provide more details on the peak selection in our data analysis.To address this, we have included a data analysis section in the Supplementary Information and have made some raw data available for reference.Methods: It is not clear how events are classified as being DNA-only or having a secondary peak.The methods section says only "a multi-step routine was used to identify the secondary peak".The workflow may be the same as for this group's recent JACS paper, but this should be explicitly outlined in this manuscript.I view this as perhaps the only glaring omission in the paper; it would be very useful to understand how events are being classified.
Please refer to Reviewer 1, Comment 1.We have added one section for data analysis in Supplementary Information.
Especially given the bandwidth here, it is likely that many subpeaks are missed.What is the dwell time distribution for the subpeaks (of different varieties)?Their apparent magnitude should also be observed to be reduced at shorter subpeak dwell times; see for example Plesa et al 2013 (https://pubs.acs.org/doi/full/10.1021/nl3042678).Are these effects observed?
For the recording we used Chimera VC100 amplifier and all data was recorded at 4Ms per second and subsequently filtered using a low-pass digital filter.Through careful evaluation of events under different filters, we ultimately chose a 1 microsecond sampling rate and a 30 kHz low-pass filter to resample and refilter the data.This selection was made to ensure a high signal-to-noise ratio while minimizing the loss of subpeaks.
In our manuscript, we did not extensively discuss the subpeak dwell time because it falls outside the focus of this particular study and introduces additional complexity.However, we sincerely appreciate the reviewer for bringing it up, as it provides an opportunity for discussion.Firstly, we would like to highlight that the phenomenon observed by Plesa et al. pertains to unbound proteins, which differs from our specific case.In our work, the translocation speed of bound proteins is primarily influenced by the DNA carrier rather than the proteins themselves.Interestingly, we do observe some phenomena that align with the reviewer's comments.
For instance, when considering the same binding position (0.2), we notice differences in subpeak dwell time and peak current depending on the entry direction.The first entry direction exhibits a longer subpeak dwell time and a higher peak current, while the second entry direction shows a shorter and smaller subpeak.This discrepancy likely arises due to the non-uniform motion of DNA molecules.Similar phenomena have been reported in Bell et al. Nature Communications volume 8, Article number: 380 (2017).However, due to differences in the experimental environment, we cannot directly apply their model to explain our results.Nevertheless, we believe this aspect would be intriguing to explore further in future studies.
The authors' statement on data accessibility seems to be that it is not readily accessible; it essentially states that the data shown in the figures is the only data they provide.How big are the raw data files that they could not be hosted on a university server?Could a subset of data, perhaps that used in key figures such as Figs. 2 or 3, be posted?This is not an essential element of the revision, but strikes me as an opportunity for greater transparency.
We agree with the reviewer and source data will uploaded as part of the resubmission.
How reproducible are the results across different nanopores?Is the data in eacy figure from single pores or aggregated across many?It is not clear whether the blockage levels, for example, are the same in different pores.Figures S5 and S6 have very similar DNA blockages, but S13 shows larger blockages -is that the pore?Are the triplicate-result binding curves all taken in the same pore?If not, are the pores nearly identical in conductance?Throughout the paper, it would be very useful to know whether these experiments are all being performed in just a few, or many, pores, as this is one of the major challenges in the nanopore field.
It should be noted that the data presented in typical traces, bar charts, or scatter plots are collected from the same nanopore.When error bars are shown, they are typically obtained from at least three different nanopores.We do not combine data across multiple pores.
The dimensions of the nanopore can vary due to pulling ambient conditions such as temperature and humidity.However, within the same set of experiments, we make an effort to use nanopores with conductance values as similar as possible to minimize variation.The experiments depicted in Figure S5/S6, which pertain to S protein detection, were conducted using nanopores pulled during the same period.On the other hand, the nanopores employed in the experiments for Figure S13 might differ since that batch was pulled at a different time.Nevertheless, our approach is specifically designed to detect the binding of target proteins by analysing sub-peaks at different fractional positions, rather than relying on peak height.This strategy effectively addresses the issue of nanopore variations.Do the results from clinical samples (for the RNA, at least) correspond to concentrations that match qRT-PCR results?It may be too late to collect these, but the data is very close to being able to verify the ability to determine viral load.This could be an important and exciting additional result to include.In a similar vein, relatively little data from the clinical samples is shown in terms of the events and scatterplots characterizing the nanopore data.The pre-clinical samples are described and a bit is shown, but the actual clinical samples only have reported statistics.Showing more of the raw data and events in SI figures would be very informative and interesting to the reader.
We compared our nanopore data with qRT-PCR results for the Delta and Omicron patient samples see (Supplementary Table 9).Unfortunately the wildtype data RT-qPCR data was not available at the time of collection.We have added some example traces and events for the clinical data, in Supporting information (Supplementary Fig. 34).
The authors are surely aware that many nanopore papers draw questions about the ease of use and requirements for pre-nanopore PCR.I suggest that they carefully re-consider some of the wording in the intro regarding the need for a new diagnostic approach without PCR, fast and on-site, simple, etc, because here they are still using PCR, and it is a highly complex detection scheme.
We agree with the reviewer's suggestion.We have made appropriate changes in the Introduction as well as the Conclusions sections.
We added these references and discussion in the Introduction (paragraph 5: line 5-7).
What is the source of the pseudovirus?Methods says just "suppliers".pseudovirus of SARS-CoV-2 was provided by Professor Ruijie Deng's lab generated using the lentiviral vector system according to their recent published work (Nat.Biomed.Eng 6, 957-967 ( 2022)).We have referenced this in the text.

FIGURES
Figures are generally very nicely laid out and designed.The extensive and thorough SI figures are particularly appreciated, as is the consistent color coding for different assays / samples that is maintained across all figures.A few minor comments and suggestions follow.The S protein concentration is 200 nM for each virus type.We added this in the legend.We also added the N protein concentration (20 nM) in the Fig. 3.
Figure 3 requires scale bars for time and current blockage for panel a.Should show an additional panel similar to 1c.Also, in d, the binning in time is too large and does not show the shape of the distribution.
The scale bars for time and current blockage have been added.We also re-binned the time data in panel d to show the shape of distribution.Fig. 2 and 3: what is the source of the S and N proteins used here?Are they from the pseudovirus?How are they purified and verified?
The S and N proteins used in Fig. 2 and 3 are not from pseudovirus but from the commercial supplier (Sino Biological).This was mentioned in the Methods section -'Translocation experiments'.We made several iterations of this figure and concluded that including tables was necessary.Without tables, the panels in (e) would be challenging to comprehend.Additionally, we made some adjustments to panel (e) to ensure that the mutant point aligns with the statistics presented in the bar chart, thereby enhancing the clarity of this figure.
Across all figures: it is good practice to include the number of events either on the figure or in the caption for histograms and scatter plots.Figures S8 and S9 have this but none of the others seem to.
The manuscript has been updated accordingly.
Reviewer #3 (Remarks to the Author): This manuscript by Ren et al. reports multiplex detection of SARS-CoV-2 antigens and RNAs using glass pipette-based solid-state nanopores and dsDNA probes.If successfully demonstrated, the potential clinical significance of this technology is clear, and the technology can be translational.However, I am afraid the current manuscript is not suitable for publication due to several major flaws in the experimental design and the data analysis as listed below.
Major Concerns: In the Introduction section, the authors failed to explain the novelty of this work in comparison to similar previous papers (JACS https://doi.org/10.1021/jacs.2c13465;Nature Communications (2021) 12:3515; Nature Nanotechnology (2023) 18:290-298).The first 2 papers are from the same group of authors.It is understandable that the first and the third papers were not cited as they are new, but the second paper was only cited without any detailed explanation.Please clearly explain the novelty of this work.
The JACS paper we published is not entirely relevant to this discussion as it explores a somewhat different topic.It focuses on the use of supercharged polypeptides as molecular carriers for nanopore detection.
In Nature Communications (2021) 12:3515, we combined single-molecule fluorescence with nanopore sensing to enable multiplexed detection.However, this approach is substantially different as it requires fluorescence labelling and an optical setup.We have also addressed this discussion in the Introduction section, specifically in Paragraph 5, lines 8-10.
Nature Nanotechnology (2023) 18:290-298, was not yet published at the time of our manuscript submission.However, their work primarily focuses on sensing viral RNA, whereas our study demonstrates the simultaneous detection of viral antigens and RNA sequences.We have included this reference in the related discussion in the Introduction section, specifically in Paragraph 5, lines 11-13.About DNA folding: There is partial folding of dsDNA probes, why not optimize the probe sequence or use smaller nanopores to try to avoid folding?
We acknowledge that optimizing the probe sequences or using smaller nanopores could potentially decrease the occurrence of DNA translocation folding events.However, we believe it would be challenging to completely eliminate all folding events by optimizing the DNA probes alone.Additionally, although employing smaller pores can reduce folding events, it would also hinder the translocation of protein-bound DNA carriers since proteins are typically larger than folded DNA.
Nonetheless, these folding events do not impact the results or their interpretability, as we have implemented a customized code to select the secondary peaks corresponding to the binding of the target, distinguishing them from DNA folding events.A newly-added section in the Supplementary Information outlines how we analyse the data and select these bound events.We believe this approach provides a more convenient and straightforward solution to address this issue.
Please double check "However, these secondary peaks, due to carrier folding, could be readily identified by their significantly smaller blockade current of secondary amplitude (90.0 ± 8.6 pA)…" Which structure causes secondary amplitude (90.0 ± 8.6 pA), S protein or folded DNA?Author claimed to demonstrate "smaller cross-section of folded DNA double-helix compared to the size of the S protein" in Figure S7a, but the figure only shows 3D structure of S protein without dimension.It is the authors responsibility to demonstrate a clear comparison of dimensions of folded DNA vs. S protein.Related to the previous comments, even with a comparison of dimensions, it is still hardly a direct evidence that the partially elevated peak is due to partially folded DNA probe.An experimental comparison between signals from a known folded dsDNA and a known unfolded dsDNA may help.
The translocation of folded DNA leads to the appearance of a secondary peak at either the beginning or the end, as demonstrated by the translocation of DNA carriers shown in Supplementary Fig. S5, S13 and S20.This phenomenon has also been documented in previous studies conducted by other research groups (Nano Lett. 10, 2493-2497(2010); Phys. Rev. E, 2005, 71, 051903;Nano Lett. 2010, 10, 8, 3163-3167;Nat Commun 10, 4473 (2019)).
The S protein's size, as determined by CryoEM in prior research (Cell, 2020, 180,281-292;Science, 367, 1260Science, 367, -1263Science, 367, (2020))), is approximately 16 nm, which is significantly larger than the cross-section of the folded DNA strand.The binding of the protein to the DNA and their subsequent translocation result in a larger secondary peak compared to those induced by DNA folding.To distinguish between these signals, we developed a customized code that identifies the target signals.This code is detailed in the Supplementary Information.
To make this comparison clear, we have compared the S protein subpeak and DNA folding level peak rather than compare the whole peak amplitude: "These secondary peaks exhibit an average amplitude of 85.2 ± 23.8 pA and a dwell time of 25 ± 18 μs.It is worth mentioning that a fraction of the observed signal events corresponds to partial folding, both in the absence and presence of S protein (Supplementary Fig. S4).However, these secondary peaks resulting from probe folding can be easily distinguished by their significantly smaller blockade current of 32.7 ± 9.9 pA, as shown in Figure 2c and Supplementary Fig. S5-6.This smaller current can be attributed to the smaller width of the folded DNA double-helix (compared to the size of the S protein, (see Supplementary Fig. S7a for the dimensions of S protein).Additionally, an experimental control was conducted using S protein alone, without the presence of molecular probes, resulting in minimal translocation events (Supplementary Fig. S7b)."About multiplex gene detection: Figure 4e shows only one calibration curve.Is this curve used for all 3 genes?It seems, in Figure 4d, that responses to different genes at the same concentration are different.In addition, how are the SD (error bars) calculated if 3 genes are sharing the same curve?
In this case, the calibration curve was constructed solely based on the response of the N gene binding ratio at a fractional position of 1.0, rather than considering all three genes.The other two binding sites (0.2 and 0.5) were designed primarily to verify the presence or absence of specific genes or mutants in the subsequent study (as depicted in Fig. 5), rather than for quantitative analysis in this particular context.We have explicitly clarified this information in the caption of Figure 4 to prevent any confusion.
Author mentioned that "For all three genes, 35-cycle PCR amplification was performed…" After reviewing the Method section, it seems RT-PCR was done to amplify the target genes in clinical samples before nanopore measurement.In this case, the nanopore experiments seems redundant, as these target genes could be directly quantified by RT-qPCR, which combines amplification and quantitative detection into one step.Even though the author argues that their method has "approximately two orders of magnitude lower LOD than gold standard RT-qPCR", clinical samples tested were all confirmed RT-qPCR positives.Without testing patients with false negative RT-qPCR, the benefit of adding many complex steps for nanopore measurement is not clear.
The integration of PCR with nanopore detection was primarily motivated by two key considerations.Firstly, the intricate secondary structure of long RNA molecules (approximately 30 kb) poses challenges for probes to access and bind to specific regions of interest.Secondly, in many real-world scenarios, unprocessed RNA concentrations can be relatively low, typically in the range of 10 -18 M (Nat.Biomed.Eng 6, 968-978 (2022)).Given that PCR is a well-established technique, we opted to combine it with nanopore detection in this study to demonstrate the proof-of-concept.However, simpler approaches like LAMP and RCA could also be employed as alternatives.
Nonetheless, the primary advantage of combining these techniques with nanopore-based detection lies in the enhanced multiplexing capabilities that allows the combined detection of multiple proteins and nucleic acids from the sample.This enables the simultaneous detection of multiple targets and the differentiation of key mutations within a single test, eliminating the need for DNA sequencing.

About preclinical test in saliva:
Author stated that "S and N protein was spiked in the saliva (with a final concentration of 20 nM)".How is the concentration calculated? in what solution/matrix?Author stated that "According to the calibration curve in Fig. 2d, we estimated the concentration of S protein to be approximately 0.5 pM, Supplementary Fig. S31."From the text related to Figure 2d, the calibration curve was not established using the same solution used in the Pseudovirus experiment, which includes "lysis and extraction buffer" etc.The calibration curve in Figure 2d cannot be used here.In addition, author only calculated LOD as 3σ above the background but did not calculate LOQ (limit of quantification).If LOQ is greater than 0.5pM, then the assay cannot quantify at 0.5pM.Again, Figure 6b results are based on a calibration curve established in a less complex solution (Figure 4e).It appears that Figure 6av is the same as Figure 2d and 3b; Figure 6bv is the same as Figure 4e.The same data should not be presented twice with different figure number references.This is not acceptable.
As described in the "Translocation experiments" section of the Methods, the pseudovirus lysate was incubated with SBA and NBA-DNA molecular probes in a buffer at a volume ratio of 1:50, equivalent to a 2% content in the buffer.For the detection of RNA in synthetic samples, only 2 µL of the PCR products were added to 200 µL of detection buffer during the validation of this method.While not identical, we believe that the 1% or 2% matrix content in the buffer does not significantly impact the detection environment.Therefore, we utilized calibration curves obtained from buffer-only experiments to estimate the abundance of target proteins or RNA in the pseudovirus or synthetic samples.
We appreciate the reviewer's concern regarding the limit of quantification (LOQ), and we have addressed this by conducting LOQ analysis for all calibration curves (Fig. 2d, 3b, and 4e).The LOQ values were calculated using a threshold of 10σ above the background signal.We have also added this information in the corresponding text to provide additional clarity.

About clinical test:
The ethics statement is inadequate.Details about how each patient sample was PCR confirmed are needed.Again, clinical samples are tested in a different buffer/matrix environment.Fortunately, there is no concentration calculation in this section, only binding ratio was presented.It is premature to conclude that the test can detect mutations reliably with only 5 clinical samples per mutation.The authors should make this clear.
We have revised the ethics statement for the use of clinical samples, providing specific details on how patient samples were confirmed.Please refer to the Methods section, specifically the Clinical Sample and Ethics Statement subsection, for the updated information.
Regarding the detection of mutations, we acknowledge that drawing reliable conclusions based on only five clinical samples would be premature.As a result, we have modified the wording of the clinical results to reflect this limitation and emphasized the need for additional clinical trials to further investigate this aspect.
In the Conclusion section, although the multiplex detection capacity is plausible, current results do not support the claim of "improved accuracy and ultra-high sensitivity".There is no comparison between this assay and benchmark methods, so improved accuracy cannot be claimed.Neither analytical sensitivity nor clinical sensitivity can be rigorously evaluated in this study.Analytical sensitivity should be assessed by calibration curves established using the same buffer as in real-world applications.Clinical sensitivity needs to be assessed using more samples and, more importantly, in parallel with clinical test results and patient information that can help reliably analyze the results (e.g.sample collection time, treatment time, symptoms, etc.)This manuscript presents a proof-of-concept demonstration of the use of nanopore technology combined with designed DNA molecular probes for multiplexed detection of viral antigens and RNA.
While our findings are promising, we recognize that there is still a considerable distance to cover before this method can be implemented in practical clinical settings.It requires extensive efforts to validate its performance in real-world scenarios.Consequently, we have revised the conclusion to reflect this perspective.The clinical claim has been downplayed, and we have explicitly stated that further clinical trials and validation are necessary before practical applications can be realised.

Minor Comments:
The author claimed that "The binding ratio for S protein increased initially and plateaued after 20 nM."However, in Figure 2d, the binding ratio of 20 nM and 200 nM showed increase.Figure S12b should be N protein?
It's N protein, Figure S12b has been revised.According to Figure 4b(i), in Figure 4c(i) the N site is positive, but the legend states negative.Similar inconsistency is also seen in Figure 4c(ii).We have revised the figure to make it easier to understand.
Figure S30 indicates ~9% events are short dwell time events suspected to be induced by background molecules.However, in the first 200 events acquired from saliva indicated in Figure S29, none of them were from background molecules.This is highly unlikely to happen statistically.
In the first 200 events, we do have several short events, however they cannot be used to calculate the binding ratio so we removed them.
"We also demonstrated sensitivity down to 1 copy of RNA per 100 μl via a preamplified step in nasal/throat swab patient samples."sounds confusing.It suggests 1 copy/100 μl sensitivity without the PCR amplification which is not the case.
Apologies for the confusion.We have reworded as "We also demonstrated sensitivity down to 1 copy of RNA per 100 µl with a 35-cycle PCR amplified step in nasal/throat swab patient samples." Please double check all references.Reference numbers are not appearing in the correct order in the text.

Thanks. References and its numbers have been checked.
There is no response to this concern: "It appears that Figure 6av is the same as Figure 2d and 3b; Figure 6bv is the same as Figure 4e.The same data should not be presented twice with different figure number references."The repetitive figures need to be removed.
The ethics statement is still inadequate.The most important information is the IRB approval.Whether this is a human study or not (only using redundant samples as materials without identifying information) should be determined by an Institutional Review Board or an equivalent committee, and the approved project number should be included in the ethic statement.
The authors are to be commended on many thorough and careful addi ons and revisions to this manuscript in response to reviewer feedback.However, one major issue that requires further clarifica on is the data analysis: While it's now clear that the authors used their Nanopore App for analysis, the procedure for iden fying secondary peaks is s ll completely elided: "a mul -step rou ne was used to iden fy the secondary peak".What are the mul ple steps?Are they different for different types of secondary peaks?Please include a descrip on of secondary peak iden fica on analysis in the Methods sec on; this en re paper is about the subpeaks, so the method for sta s cally iden fying and characterizing them should be explained thoroughly and clearly.
We appreciate the feedback, and as a result, we have expanded upon the explana on of our secondary peak analysis methodology in the Supplementary Informa on, as well as in the sec ons dedicated to data analysis and subpeak analysis.These sec ons now provide a more detailed account of the threshold employed for defining subpeaks in this paper.
Note that in the SI Note "Data Analysis", the analysis step #s do not line up with the #s in the main text (example: main text analysis step 2 is "Track and subtract the baseline" while SI Note step 2 is "Resample and Filter Trace").Moreover, in the SI Note "Data Analysis" the only men on that is made of sub-peak finding is currently in SI step #6 (the SECOND SI step #6, "Subpeak Analysis") and says just that "By clicking the 'Auto Subpeaks' analysis bu on, the subpeak details for all events containing subpeaks will be compiled and listed.Thresholds are employed to dis nguish genuine posi ve events from poten al false posi ves caused by folding signals."This does not explain how the analysis is done -what thresholds and sta s cal tests are used to iden fy and screen poten al subpeaks.
Thank you for point this out, we have revised the main text to align the steps with the SI.We included an explana on in SI of what thresholds and sta s cal tests are used to iden fy poten al subpeaks.
Is the figure on p. 46 of the SI (no cap on) intended to be an addi onal SI figure related to the subpeak finding?If so, it needs a cap on and explana on, and a clear reference within the SI note.It appears to show examples of subpeaks and folds rather than explaining their sta s cal determina on.And, given that it is showing a data set in which only 7 events have subpeaks, what is the reader meant to learn from the histograms of those seven events into five bins ("subpeak current")?
We added the cap on.We also removed the histograms of the subpeak here.
On a related note, the authors' asser on in the response to reviewers that "The width or height of the subpeaks does not impact the conclusions drawn from the data because we rely on the ra o of bound to unbound events to dis nguish posi ve and nega ve events" does not make sense since they (presumably; methods unclear) rely on width and height of subpeaks to differen ate folded states from bound protein subpeaks.
We regret any confusion our ini al response may have caused.To clarify, following the selec on of posi ve events, we observed slight varia ons in the subpeak's width and height.This finding aligns with the answer we previously provided concerning transloca on speed.However, these varia on in the subpeak parameters, are significantly smaller than and dis nct from the DNA folding signal, and in effect do not interfere with the subpeak ra o calcula on.Our subpeak selec on methodology leverages mul ple parameters, including frac onal posi on and either subpeak width or height.These parameters aid in dis nguishing poten al signals that differ from the DNA folding signal.
We have therefore included a revised suppor ng figure S17 (now as Supplementary Fig. 8 due to the order changes) and cap on along with the addi on of text in the manuscript (page 5 paragraph 2 for S protein and page 7 paragraph 2 for N protein) to clarify how we dealt with folded events.We also included and addi onal Supplementary Fig. 17 (now as Supplementary Fig. 8), that shows detec on and subpeak analysis of S and N protein and folded DNA probe.The frac onal peak posi on histograms are used for the discrimina on between folded DNA probe and bound N protein.
The authors also state that they "have made some raw data available for reference" but I do not see this in the review materials -only figure data files of processed data.
Unfortunately for the previous submissions we could not upload the data with the NPG repositories available for review material, due to size limita on (the data are 31.4gigabytes).Now we have uploaded example raw traces to ZENODO general database (h ps://zenodo.org/record/8143395)Reviewer #3 (Remarks to the Author): Regre ably, some of the concerns are not successfully addressed.In general, the authors answered most ques ons only in the response le er, but failed to incorporate the explana ons into the paper to improve scien fic rigor.
The authors explained the issues of par al folding of DNA probe sufficiently.Please implement this explana on in the manuscript so the readers may understand why addi onal op miza on to eliminate folding is not necessary as it does not impact result in interpretability.
Based on the previous revisions we added a discussion on the issues of par al folding on Page 5.However, we agree it was not as clearly presented as we would have liked.We have therefore included a revised Supplementary Figure S8 and cap on along with the addi on of text in the manuscript (page 5 paragraph 2 for S protein and page 7 paragraph 2 for N protein) to clarify how we dealt with folded events.
In the mul plex gene detec on sec on, the current text implies simultaneous quan fica on which is not achieved.Please clarified, in the main text, that the calibra on curve was constructed for N gene, and that the other two binding sites were used to verify the presence of the other 2 genes.This is only vaguely men oned in the figure cap on in the current manuscript.
We have added further clarifica on on this subject in the main text to avoid confusion (highlighted in Page 9, Paragraph 2).
Please add the ra onale of the need of PCR to the manuscript.This is only explained in the response le er but not incorporated into the paper.
We have amended the explana on in the main text (Page 8, Paragraph 2).
Please acknowledge the difference between the buffers/matrices used for calibra on curve and the other experiments in the manuscript.The lack of this informa on will cause confusion and reproducibility for poten al readers.
We added a clarifica on on this difference to avoid confusion (highlighted in the last paragraph in Page 13).
There is no response to this concern: "It appears that Figure 6av is the same as Figure 2d and 3b; Figure 6bv is the same as Figure 4e.The same data should not be presented with different figure number references."The repe ve figures need to be removed.
We have removed the Figure 6a(v) and Figure 6b(v) and referred to Figure 2d/3b or Figure 4e.
The ethics statement is s ll inadequate.The most important informa on is the IRB approval.Whether this is a human study or not (only using redundant samples as materials without iden fying informa on) should be determined by an Ins tu onal Review Board or an equivalent commi ee, and the approved project number should be included in the ethic statement.
We amended this ethic statement as follows: "The viral samples were provided from the Imperial College London tes ng scheme and were from fully anonymised, redundant samples (i.e., samples le over a er tes ng) and retained for assay development, quality assurance, and valida on.The sequencing was part of the Imperial College London's response to ensure that new variants were detected and to detect otherwise unexplained clusters.The consent to providing the sample and to the tes ng of the sample was provided at test booking through an online process.The use of the virus was in accordance with RCPath guidelines." In the UK an IRB would not be relevant for use of these samples.Guidance from the Royal College of Pathologists, which is relevant in our case, state that le -over sera or plasma should be stored for as long as prac cable, to provide an array of material for future research and disease surveillance purposes.While long-term storage may be imprac cal in many se ngs, virology centres and laboratories involved rou nely in public health ac vi es should retain sera for a minimum of one year to facilitate 'look-back' exercises, iden fica on of emerging infec ons and vaccine programme monitoring.Samples that do not contain human cells are not regulated as human ssue by the Human Tissue Act, although ethical constraints on appropriate storage and use nevertheless apply.
My thanks to the authors for their updates.In my view, the description of subpeak finding and identification is still inadequate, and has barely been updated as far as I can tell.I have included some specific suggestions below -again -in case the authors see fit to add this information.
As far as I can tell, only vague additional language was included to describe finding the subpeaks.Your methods section still says only that "a multi-step routine was used to identify the subpeaks".What routine?On what basis are subpeaks identified?WHAT are the multiple steps???If your next sentence "Parameters such as secondary peak amplitude, dwell time, and fractional position were determined" is related to identifying subpeaks as protein or DNA, you need to be more specific.What ranges of these values, or what thresholds, were used to differentiate protein and folded DNA?In what experiments were additional separating parameters needed, and did the cutoffs vary between experiments?
Similarly, in the SI you still say only "By clicking the 'Auto Subpeaks' button, relevant information..." What does the 'Auto Subpeak' button DO, statistically?You say "thresholds could be employed"... but WERE they employed?If so, what thresholds?Were the thresholds different across data sets?If so, why?
My thanks to the authors for their updates.In my view, the descrip�on of subpeak finding and iden�fica�on is s�ll inadequate, and has barely been updated as far as I can tell.I have included some specific sugges�ons below -again -in case the authors see fit to add this informa�on.As far as I can tell, only vague addi�onal language was included to describe finding the subpeaks.Your methods sec�on s�ll says only that "a mul�-step rou�ne was used to iden�fy the subpeaks".What rou�ne?On what basis are subpeaks iden�fied?WHAT are the mul�ple steps???If your next sentence "Parameters such as secondary peak amplitude, dwell �me, and frac�onal posi�on were determined" is related to iden�fying subpeaks as protein or DNA, you need to be more specific.What ranges of these values, or what thresholds, were used to differen�ate protein and folded DNA?In what experiments were addi�onal separa�ng parameters needed, and did the cut-offs vary between experiments?
Similarly, in the SI you s�ll say only "By clicking the 'Auto Subpeaks' buton, relevant informa�on..." What does the 'Auto Subpeak' buton DO, sta�s�cally?You say "thresholds could be employed"... but WERE they employed?If so, what thresholds?Were the thresholds different across data sets?If so, why?
Thank you for the feedback on our paper.We have taken into account your concerns regarding the descrip�on of subpeak finding and iden�fica�on.Here is a list of the changes that address the points raised: 1. Clarifica�ons made in the previous version: contrary to the percep�on, we have made significant amendments in our previous revision.For instance: a.The SI -data analysis sec�on now provides a comprehensive account of how the peaks are determined.b.We updated the SI figure S8 to emphasise the subpeak analysis.c.The SI text has undergone significant modifica�ons to incorporate sta�s�cs related to the subpeaks.
In this version, we have included an even more detailed discussion that can be found in Supplementary Fig. 36-37 2. Descrip�on of 'Mul�-Step Rou�ne': a.We acknowledge that the term "mul�-step rou�ne" was ambiguous.We meant it to encompass all stages of data analysis.We removed the term and described the analysis step-by-step.b.The subpeak iden�fica�on relies on several parameters like peak height, dwell �me, and frac�onal posi�on.c.The details and all the parameters for each target experiment are described below and in the Data Analysis sec�on in Supplementary Informa�on (Supplementary Fig. 36-37).
3. Detailing the thresholds for each target: a.We have incorporated descrip�ve paragraphs and figures that clearly lay out the criteria for subpeak selec�on.This addresses: i.How each parameter is employed.
ii.The suppor�ng schema�c of the parameters used to iden�fy posi�ve subpeaks.
iii.The threshold values for each parameter when choosing various signals.b.An in-depth explana�on of thresholds applied for specific detec�ons (S protein, N protein) has been included, explaining the ra�onale behind the chosen thresholds.
In the updated SI with addi�onal discussion and supplementary Fig. 36, we illustrate what parameters and thresholds are used and how they are defined: The last step of the data analysis involves extrac�ng the subpeak informa�on.By clicking the 'Auto Subpeaks' buton, relevant informa�on such as subpeak amplitude, subpeak dwell �me, the frac�onal posi�on of the subpeak, frac�onal width of the subpeak, number of subpeaks, loca�on of subpeaks could be extracted directly from the CUSUM fits which has been illustrated in Step 5. Due to DNA folding, thresholds could be employed to dis�nguish between genuine posi�ve events and poten�al false posi�ves.The frac�onal posi�on is o�en used to isolate events associated with subpeaks origina�ng at a defined loca�on.As the subpeak width for the protein is typically smaller than that of folded DNA, this can also be used as a threshold.Peak amplitude can also be used as a discriminator (Supplementary Fig. 36).
a.The schema�c of the key parameter for select posi�ve events in this experiment.b.The threshold value for each parameter in the selec�on of different signals.
Then, we illustrate the detailed thresholds for each target and the reason for choosing them: In detail, for S protein detec�on, the threshold for the normalised subpeak posi�on (The 'Frac�onal Peak' in the Nanopore App interface) has been set to 0.1 ± 0.1 and 0.9 ± 0.1 as the S protein would bound to the SBA located at the end of the molecular carrier.To discriminate the S protein signal and par�ally folded signal, which is observed at the events' beginning or end, another two thresholds, subpeak width and subpeak/DNA ra�o, have been applied.For par�al DNA folding, the subpeak always shows a wide rectangular shape with a longer dwell �me (0.447 ± 0.231 ms) and is very unlikely to show a narrow, sharp spike due to the limita�on with the persistence length (> 50 nm).However, the S protein binding results in a short dwell �me and high current amplitude secondary peak due to the rela�vely larger size of the S1 protein compared to the folded DNA.Here, the threshold for subpeak width was set to <0.2 ms to select the events with narrow subpeaks where the threshold for subpeak/DNA ra�o was set to >1.5 as the folded DNA subpeak/unfolded DNA level was 0.662 ± 0.027 (Supplementary Fig. 36).It should be noted that when the transloca�on event exhibited both folding and protein binding signals, these events were individually examined to verify that all binding events were classified correctly.
The representa�ve DNA folded events that can be filtered out, the S protein bound events that can be selected automa�cally, and the S protein bound events that contain DNA folding, which requires a manual check, were listed in Supplementary Fig. 37.
The threshold for the normalised subpeak posi�on for N protein detec�on has been set to 0.5 ± 0.2.This is in good agreement with the placement of N protein binding site, which is designed to have a frac�on posi�on of 0.48 based on the sequence on the molecular carrier.Subpeak amplitude was not used to discriminate between events due to the similarity in amplitude between folded and proteinbound events.The observa�on of DNA knots in the middle of the transloca�on event is uncommon (< 0.1 %), and hence, all protein-bound events could be isolated based on normalised peak posi�on and subpeak width (<0.3 ms).Similarly, all folded events were cross-checked manually to ensure all binding events were counted as par�al folding will lead to a slight shi� in the frac�onal posi�on.For RNA 1.0 site, we use a similar strategy as this is N amplicon site, and it is responsible for confirma�on of the posi�ve signal and RNA quan�fica�on.Therefore, in addi�on to applying a 0.1 ± 0.1 and 0.9 ± 0.1 normalised subpeak threshold and <0.2 ms subpeak width threshold, the events were always checked manually to ensure accurate coun�ng.
4. It should be noted that across different nanopores, the threshold values remain the same; the reason is: a. normalised subpeak posi�on would not affected significantly by the nanopores with different sizes/shapes; b. the dwell �me difference in nanopores with a size in a very narrow range is very subtle; c.The change in peak amplitude values caused by different nanopores is most evident.
However, we used the ra�o of Subpeak to DNA level as a threshold and u�lised normalisa�on to eliminate this effect.In Supplementary Fig. 37, we have shown 32 events collected using different nanopores and classified using the same thresholds.
5. We also added one figure showing that using the thresholds, the DNA false posi�ve signal can be filtered out, the protein-bound real posi�ve signal can be selected automa�cally, and the folded DNA-bound protein signal needs to be checked manually.We also illustrated that the cut-off values did not vary in different devices and explained the reason above.
Supplementary Fig. 37 | Example events that have been isolated using the thresholds.
U�lising specific thresholds, we can dis�nguish and separate events with relevant subpeaks.Signals from molecular carriers, regardless of their folded or unfolded state, are excluded (indicated in the red box).Events that sa�sfy all threshold criteria for subpeak informa�on are iden�fied as poten�al posi�ve signals (highlighted in the green box).For events exhibi�ng both folding and protein-binding indica�ons, individual assessments were made to ensure proper classifica�on of all binding instances (shown in the yellow box).The events we listed here are all recorded with different nanopores.
Clicking the 'Auto Subpeaks' buton will classify individual events based on the informa�on obtained from the CUSUM fit.A CUSUM fit (represented by the red dashed line) is employed for each event.
Different molecules, such as DNA protein, cause different disrup�ons in the current.The resul�ng signal has "steps" corresponding to these different molecules' size and charge.The CUSUM algorithm is used to detect these "steps".Once these significant changes or steps are detected, the con�nuous signal can be segmented into "events", each event corresponding to a par�cular DNA, protein or DNA/proteinbound complex.
In summary, we have made deliberate efforts to provide a more in-depth explana�on of our methodology, supported by revised supplementary figures.We hope these clarifica�ons and addi�ons address the reviewer's concerns, and we appreciate the opportunity to enhance the clarity and thoroughness of our work.

REVIEWERS' COMMENTS
Reviewer #2 (Remarks to the Author): My thanks to the authors for providing the details of subpeak analysis.I will be delighted to see this publication in Nature Communications in the near future.

Fig 4 .
Fig 4. A very useful SI figure might be an all-subpeak histogram related to the data split up into the three panels of (iii).

Fig 5 :
Fig 5: c and d might be better as tables.A and b seem a bit overkill.E is really the most important part of the figure.But, my comment is more about style than substance; OK if authors prefer to leave it.
The supplementary Figure2caption has been revised.5.It is recommended to give the value of Kd for the SBA in the main text as the NBA does.We calculated the values as Kd for NBA = 0.73 nM Kd for SBA = 9.86 nM We have updated Fig.S1 and Fig. S10 captions with the Kd value.

Figure 2 :
Figure 2: what are the protein concentrations for each sample in 2e?

Fig 4 .
Fig 4. A very useful SI figure might be an all-subpeak histogram related to the data split up into the three panels of (iii).We have added a new SI figure to support this.(Supplementary Fig. 35?)

Figure
Figure 5e is confusing.I suggest matching the target of each site (colored blocks on the top) to its fractional position.For example, N at last etc.