This article uses computational methods to determine possible proteins that would be able to successfully bind trimethoprim within blood serum. As stated by the paper, Trimethoprim can cause renal problems and even induce allergic reactions. The authors purport that if a protein can be found that can binds to trimethoprim, it could “provide advantage in terms of time and cost” by developing an assay to determine the concentration of trimethoprim. However, the authors do not state what exact costs they are reducing with this method, leaving some ambiguity to the possible benefits to this research. The authors also state that the ability to determine trimethoprim concentration within serum would allow further study of the pharmacokinetics and efficacy of said drug. The claims by this paper do have some importance as current methods of detection rely on costly instruments such as HPLC, as stated below.⁶

Other methods of determining Trimethoprim in serum have focused on the use of HPLC.⁶ It appears that the method proposed by the author is novel for the field.

In the discussion about the docking study within the paper,  authors point to previous research that shows that computationally determined binding energies correlate with empirically determined ones.¹ The article cited to back up this claim however, is unrelated to the application currently being used for docking, instead referring to a docking application from 1997 that uses a force field type equation rather than an empirical/knowledge based function as seen in Autodock Vina.² While Autodock Vina is a very popular docking application, its ability to correlate its scores with the binding energies seen within the literature has yet to be proven.³ Thus, relying on this methodology to determine which protein would best bind with trimethoprim may produce inaccurate results. 

The conclusion that Pseudomonas putida shows the best dihydrofolate reductase for trimethoprim binding does rise naturally from the data provided from the docking studies. With a docking score of -14.7 kcal/mol, the stated protein outpaces its competitors by -.9 kcal/mol. However, with the mentioned issues with scoring functions, the experimental validity of these scores will have to be determined.

With the protocol provided, it is unknown if there are any deviations resulting in the data. One way to test the validity of the method would be to procure proteins for which trimethoprim scores poorly in, and then used the same protocol with those proteins. If the score from Vina is worse, then it shows that the method the authors are using is correctly predicitng some level of binding affintiy with the proteins tested.

As stated before, the ability of Autodock Vina, or any other docking program, to determine experimental binding energies has not been demonstrated.³ However, the methods used to generate the protein structures from previously known crystal structures through homology modeling has been shown to be accurate.^4,5

While the methods presented in the paper have mixed validity for the paper, the explanation of methods leave one wanting for more elaboration. In the 3D model, active site characterization, and docking ligand sections within the methods, no parameters are given for any applications used for the generation of these crystal structures, determination of active site, or the docking of trimethoprim into the modeled proteins. The structure of the methods section is well put together however, with good organization of the different steps taken to get the final set of docking data with trimethoprim.

It is unknown whether the scores generated by Autodock Vina are the true “best” poses. It has been shown, after this paper was published, that Vina is only able to produce the best ligand pose 48.5% of the time.³ If these docking studies were only performed once, it is likely that the best pose was not generated.

In terms of the experimental design, one possible flaw would be the use of different templates for the homology modeling of the proteins. Choosing to model each protein based on the template with the best fit can lead to bias, and exclude possible models that may fit better other templates. To improve this method, authors may want to consider running each of their sequences with each template to show that the modeling is best with the template they propose. The amount of proteins used for comparison regarding trimethoprim’s binding energy makes it clear that there are large differences regarding the active site of these relatively similar proteins.  The docking studies also have a reliability problem, as it is unclear whether Autodock Vina is able to reproduce the crystal structure pose of the original template used for homology modeling, if Vina is unable to perform that task, then the application is unsuited for this task.

Obtaining the data for these suggested experiments would not be overly exhaustive, as the method for docking and homology modeling has already been created for this system.

This paper contains some flaws that need to be fixed before publication. The previously mentioned flaws with autodock vina and with the limited homology modeling should be resolved.

Also, while the independent variable of the experiment is the protein structure docked with trimethoprim, the authors leave out any discussion of why certain proteins bind better with trimethoprim. While there is quantitative data from the computations performed by Vina, there is no qualitative discussion about the active sites of the proteins and why they have higher affinity for the drug target. This could be easily shown with the previously mentioned docking images showing the intermolecular forces in the protein/ligand complex. Without this discussion, it appears that the top protein wins merely by luck, rather than any empirical reason.

1.  Ausiello, G., Cesareni, G. & Helmer-Citterich, M. ESCHER: a new docking procedure applied to the reconstruction of protein tertiary structure. Proteins Structure Function and Genetics 28, 556–567 (1997).

2.  Trott, O. & Olson, A. J. AutoDock Vina: improving the speed and accuracy of docking with a new scoring function, efficient optimization, and multithreading. J Comput Chem 31, 455–461 (2010).

3.  Wang, Z. et al. Comprehensive evaluation of ten docking programs on a diverse set of protein-ligand complexes: the prediction accuracy of sampling power and scoring power. Phys Chem Chem Phys 18, 12964–12975 (2016).

4.  Sali, A., Potterton, L., Yuan, F., van Vlijmen, H. & Karplus, M. Evaluation of comparative protein modeling by MODELLER. Proteins 23, 318–326 (1995).

5.  Schwede, T., Kopp, J., Guex, N. & Peitsch, M. C. SWISS-MODEL: An automated protein homology-modeling server. Nucleic Acids Res 31, 3381–3385 (2003).

6. Hung, C.T. and Perrier, D.G., 1985. Determination of trimethoprim and sulphamethoxazole in serum by reversed-phase and ion pair HPLC. Journal of liquid chromatography, 8(3), pp.521-536.

.

I have performed several docking studies on multiple proteins with Autodock Vina.