Albumin-Binding Domains in Therapeutic Protein Engineering: A Structural and Computational Perspective on Rational Design

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

In the review paper entitled “Albumin-Binding Domains in Therapeutic Protein Engineering: A Structural and Computational Perspective on Rational Design” by M. J. Argyle, D. Chipman, A. C. Woolley, B. C. Bundy, and D. D. Corte, the authors present a comprehensive overview in which albumin is considered a valuable target in therapeutic protein engineering. The manuscript emphasizes the analysis of albumin-binding domain (ABD) fusion protein design, integrating insights from structural biology and computational prediction.

Parts 2 and 3, “Albumin-Binding Domains: A Structural Classification” and “Linker Architectures in Protein Fusions,” are particularly interesting. These sections provide a clear and informative overview of human serum albumin in complex with various albumin-binding ligands targeting different domains. The text is well written and supported by an appropriate number of key references.

Nevertheless, there are several points that I would like to address:

1. The review does not include a discussion of albumin complexes with inorganic nanoparticles. If the authors intentionally chose to narrow the scope of the manuscript, it would be helpful to explicitly state this and briefly acknowledge that substantial research exists in this area, providing references to key reviews without detailed discussion.
2. The influence of salt concentration and buffer conditions on albumin complex formation is not mentioned. This factor can be particularly important, especially given the authors’ recommendation at the end of the review to evaluate systems using biological assays.
3. Minor comment: Two sentences appear to be identical in lines 194–196.

Part 4, “Structural Comparison: Experimental and Computational Approaches,” is somewhat confusing. After a concise overview of what is known from X-ray crystallography and NMR methods (Section 4.1), the authors move on to discuss the advantages and challenges of using computational tools such as AlphaFold for protein–protein complex prediction. The manuscript suggests that different scoring metrics of predicted linker complexes may justify the use of these structures for subsequent experimental validation.

While AlphaFold has indeed provided a significant advance in the structural study of biomolecular complexes, intensive research is still ongoing to fully understand its applications and limitations. In my opinion, this field remains at a relatively early stage. There is an emerging consensus that AlphaFold-generated starting structures or conformational ensembles can be cautiously considered as inputs for further experimental validation, but only with appropriate care.

With this in mind, the message conveyed in the section 4.5 “Integrating Computational Optimization and Experimental Approaches,” particularly with respect to industrial applications, appears somewhat inaccurate or overstated. In Scheme Figure 4, a critical step seems to be missing between Structure Prediction (Step 3) and Expression & Purification followed by biological assays, including animal studies. Step 4 represents a highly resource-intensive and costly process, making it impractical to proceed directly based on predictions from systems that are not yet fully reliable.

Even in pharmaceutical research, small-molecule hits identified through well-established virtual screening approaches typically undergo careful secondary evaluation before entering synthesis and testing cycles. Therefore, I recommend that the authors reconsider and revise Section 4.5, particularly the recommendations related to industrial and applied use.

Overall, this manuscript provides a valuable and timely overview of albumin-binding domains in therapeutic protein engineering. The review is well structured and contains several strong sections of clear relevance to the field. I would recommend the manuscript for publication after the authors have addressed the points raised above and revised the manuscript accordingly, as these revisions would significantly improve clarity, balance, and the accuracy of the conclusions, particularly with respect to computational approaches and their practical application.

Author Response

Please see the attachement

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

In the manuscript "Albumin-Binding Domains in Therapeutic Protein Engineering: A Structural and Computational Perspective on Rational
Design" by Argyle et al. the authors discuss the use of albumin-binding domains in fusion proteins to improve the lifetime (i.e. clearance), thermostability, and binding affinity of the target protein. The authors also then describe their development of three-helix bundle albumin binding domains. Additionally, the authors describe the increase in circulating lifetimes for proteins that are fused to albumin. Overall the paper is well-written and suitable for publication once a few minor points are properly addressed. 

Minor points:
 - "exhibits a half-life of a measly 2 minutes" (line 46) is colloquial phrasing and should be rephrased. I understand that this helps to differentiate the paper from AI generated text but its not appropriate in a professional manuscript. 
 
 - the benefits in circulating lifetime due to albumin are quite interesting although this does not get mentioned in the abstract. Adding a sentence to the abstract about the notable benefits of albumin-fusion would probably enhance the impact of this manuscript.

- figure 4 is a bit low resolution in the current pdf version. This makes reading the dark grey text on a light grey background in the "timeline" section difficult to read, for example.

- the paragraph discussing the benefits of AI structure prediction (lines 104 - 118) needs to be better supported or edited to a point that is more solidly supported or linked better to the later text. For example, while AlphaFold has more than 200 million structure predictions, but many of them are low quality predictions and the models contain a sizeable proportion of poorly or un-predicted structure. Additionally, the ability of the linker to provide adequate spatial separation can be done, generally successfully, without AI. Especially if the protein is generally predicted to be highly similar to a protein with an already known structure. The authors should be more academically neutral to the benefits of AI.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

I am satisfied with the revised version of the manuscript. Almost all of my comments have been adequately addressed, and the suggested corrections have been implemented. Therefore, I recommend that the manuscript be accepted for publication and believe that this review will be useful for a broad readers.