Feature-Based Growth Curve Classification Enables Efficient Phage Discrimination

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The article ‘Feature-Based Growth Curve Classification Enables Efficient Phage Discrimination’ proposes a statistical evaluation of phage-growth curves and distinction of phages based on chosen few statistical descriptors. The authors present results for both model phages and isolates, complementing their analysis with leave-one-species-out validation and biologically significant analysis of phage growth curves from single plaques.

The main value of the work lies in the improvement of phage identification techniques. The reviewer agree that this field needs improvement. Many phage evaluation techniques still depend on very basic, labour-intensive techniques of limited output. Such techniques limit general development of phage-based therapies very needed in face of approaching antibiotic resistance crisis.

Typically, a mathematical model explaining and predicting phage curves should be proposed, however this task in biological setups is often extremally difficult if not impossible. Results of such works often lead to overcomplication of the pipeline limiting its effectiveness. Authors attempt to simplify a biological sample to 7 easier-to-use datapoints. The reviewer recognizes that and agrees with rationale of authors. The improvement over existing tools is also proven by authors and reviewer agrees that authors validated their pipeline successfully fulfilling necessity of novelty and scientific output for scientific manuscript. However, the improvement is minor, and limitations are clearly seen in assessments of varying MOI or sewage isolated phages. Authors understand it and sufficiently discuss limitations and provide statistical data even when putting their pipeline in the poor context. The reviewer notices such honesty and appreciates.

Minor improvements should be done:

Post-hoc estimation of MOI in 3.4 would complement the analysis. If authors have this data, it should be added?

Line 32: ...822 million (annully?) – please specify

Line 377: after excluding one T6 isolate – mark this isolate in supplementary table.

Table S6 contain data for curves of T7 phages isolated from plaques, but there is no citation for it in 3.4 section. Similar situation is for table S7 and section 3.5. – authors need to make sure that citations are correct. No in-text citation for Table S5 or S7 either. Correct me if I am wrong, but I believe that there is policy that all supplementary figures, tables or data need to have in-text citation.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

This manuscript proposes an upstream screening method that utilizes co-culture growth curve analysis to rapidly screen phage isolates during the early separation process. They extracted seven biologically significant features from the bacterial growth curve, which can reflect the lysis kinetics, lysis efficiency, and post-lysis dynamic changes, and applied an unsupervised clustering algorithm for phage identification. This method reduced the number of candidate substances to two-thirds (from 21 to 7 isomers) while maintaining complete species coverage, thereby providing an efficient and scalable screening tool, reducing the workload of downstream analysis, and accelerating the discovery of novel therapeutic phages for clinical application. This research is an innovative work. However, before acceptance for publication, the following issues need to be discussed with the authors:

1. What are the differences between the two host bacteria, NBRC13168 and NBRC13898? Why were they selected for the T phage and environmental phage experiments respectively?

2. On page 7, the color bar range of "Spearman Correlation" is from -1 to 1, while the values in the figure are mostly displayed between -0.5 and 0.5. Why?

3. In the PCA graph, species are represented by shape and clustering results by color. However, in the legend of Figure 2a, "Species" and "Cluster" can be further optimized to be more dispersed to avoid overlap and facilitate reading.

4. On page 10, the problem in Figure 3a is similar to that in Figure 2a, especially the legend in the upper right corner is too overlapping. It can be referenced for correction to disperse the legend to avoid overlap and improve aesthetic clarity.

5. On page 4, 2.5.2: The feature "OD at bottom" is defined as the OD value when the lysis rate drops to 10% of the maximum rate. How was this 10% threshold determined? It can be briefly noted to make it clearer.

6. On page 5, 2.6.2: In LOSOCV, when a species is excluded, how is the number of clusters K re-determined? Is it using the same rule or fixed at 6?

7. The calculation process or formula of the Sampling score can be supplemented to facilitate readers' easier reading and understanding.

8. On page 6, 2.8: For two sequencing data, both Flye and SPAdes software were used for assembly. Why were different assembly tools selected? Can the consistency of the results after processing the data with the two software be guaranteed?

9. On page 11, 3.5: The clustering ARI of environmental phages is low (0.196), but the sampling score is 1.0. Does this mean that GC7 can effectively distinguish species but cannot perform fine clustering within species? Please clarify.

10. In Figure 5g-i of P11: The article hypothesized that the lack of species-specific patterns in the time distribution from the bottom to the increase in OD, peak count, and decrease magnitude may affect the clustering accuracy of phages. The fact that certain features show no significant difference between genomic-defined species indicates that this may be the main reason for the low ARI. Does this mean that for environmental phages, additional or different features need to be extracted?

11. Provide a more specific discussion on the scalability of GC7 on other bacterial species.

 

 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

The authors present a potentially very useful method to quickly differentiate between isolated phages based on their infected host’s growth characteristics. Especially appreciated is the effort to benchmark their system not only on MOI controlled lysates but at different MOIs and even based on plaques alone.

I would love to implement this method in my own research, and I’m sure so would many others, especially since it is very low effort/cost/technology, plus easily scalable, so could be used in many different research settings. However, this is only possible if the manuscript allows for facile replication of the methods. In the current state, it absolutely does not, and revisions are needed.

Main issues:

1. I) Starting from a growth curve, the authors speak of a pipeline that includes:

• Blank substraction – makes sense
• Noise reduction via a scipy library function – not clear which command but probably something that can be found in the scipy library (which is not cited)
• Setting minimum t=0 as mimum OD – makes sense

Further:

• Counting peaks via scipy, specific command given – makes sense

From then on it reads more like an explanation of the methods that would be found in the results section than something that can be repeated. What programs were used to do any of these calculations? Any citations for these algorithms/formulas?

At the very minimum, the authors should provide the actual code used to go from beginning to the end of the analysis (github page?) for their provided raw data, allowing the reader to replicate their analysis and adapt it to their own use.

Ideally, the authors should provide a single script to analyse the user’s growth curve and perform analysis automatically. They might deem this outside of the scope of the article, I would certainly not make this specifically stand in the way of publication, but it would ensure that other people actually use their method.

1. II) Additionally, since most initial discerning of different phages is based on plaque morphology, having pictures of the plaques formed by all phages in this study would be an extremely helpful demonstration of the use of this system. Otherwise why not just trust our eyes?

III) An ANI graph for all the T-even phages and their taxonomic classification would be helpful. T3 and T7, which their method is unable to discern, are in fact very closely related, so this would strengthen their results.

Line-by-line:

105 – needs clarification. Picked plaques – how? Cut out, touched with a pipette tip etc. Was there a lot of plaque size variation for same phage that could account for different results? Hence the request to have pictures of plaques.

107 – needs clarification. How much filtrate was added to how much bacterial culture (volumes)

120: Starting from here, it is becoming very complicated to follow the methods in a way that would lend itself to replication.

438: phage or phase?

Figure 1: It would be very helpful to have a visualization of these measurements on an example growth curve (either as a panel here or a supplementary figure). Legend – indicate n (looks like 3, just making sure)

Figures 2-5: y axes should be normalized across figures for the same measurement –> same ARI scale across figure etc. Otherwise this is somewhat misleadingly showing seemingly good scores for one method when in reality it is just the least bad (compare ARI for figure 5c and 2b)

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

I have no further remarks. I wish authors all the best.

 

Best regards!

Reviewer 3 Report

Comments and Suggestions for Authors

The authors have addressed my concerns, ready to publish!