Comparative In Silico Analysis and Functional Characterization of TANK-Binding Kinase 1–Binding Protein 1

Introduction

The prediction of the 3-dimensional (3D) structure of proteins from amino acid sequence has been very challenging in computational biophysics for decades. Recently, studies have discussed the inverse problem in which an amino acid sequence designed will fold into a specified 3D structure. This phenomenon has attracted the attention towards the rational engineering of proteins which is useful in biotechnology and pharmaceutical sciences. Previously reported studies show the methods for the prediction and design of protein structures and now they are greatly advanced. New algorithms for protein designing folds and protein-protein interactions have been used to engineer the novel assemblies and to design fluorescent proteins with improved properties, as well as signalling proteins which have the improved therapeutic potential. ¹

Proteins are important components of life, and by understanding the 3D structure, we will be able to understand their physical, biological, and biochemical properties very efficiently. About 100 000 3D structures of novel proteins have been predicted till date.^2-5 But this is considered as a very small fraction compared with the enormously known protein sequences.^6,7

TBK1-binding protein 1 is present in Homo sapiens having a sequence of 615 aminoacids with a primary accession number A7MCY6 in UniProt. ⁸ TBK1-binding protein 1 is considered the major regulator of interferon type 1 and it interacts with different types of proteins such as NAP1, TANK, and SINTBAD. This protein is called the adapter protein because it has several binding sites through which it interacts with the other proteins such as TBK1 and IKBKE and performs their functions in antiviral innate immunity. ⁹ It is present in lungs, small intestine, leuckocytes, heart placenta, muscle, kidney, lower level of thymus, and brain. It is encoded by gene TBK1 present in Homo sapiens. ¹⁰ It interacts with different proteins; it mostly forms homodimers but sometimes forms heterodimers with NAP1. ¹¹

Determination of 3D structure of a single protein is a laborious method of about a month to years in the past and different experimental techniques such as x-ray crystallography were used which were expensive and time-consuming. To fill the gap, there is the need of fast and accurate computational methods for 3D structural prediction. Due to the fastest progress in structural bioinformatics, the fast and free-of-cost techniques have been introduced so that everyone can easily access the notable servers such as I-TASSER, AlphaFold, and Robetta. ¹² I-TASSER is an approach used for the 3D structures of proteins and structure-based functions. First, it identifies the structural templates for protein from PDB (Protein Data Bank) by multiple threading approach and models are constructed by iterative template-based fragment assembly stimulations. ¹³

AlphaFold is an artificial intelligence system that helps in determining a protein’s 3D structure from its amino acid sequence. It has a competitive accuracy. ¹⁴

Robetta is an online server providing automated protein structure prediction and analysis. ¹⁵ Robetta was used for the determination of 3D structure, and the models were validated using the SAVESv6.0 (PROCHECK) server. ¹⁶

Methodology

Building of protein model and its evaluation

The 3D structure of the targeted protein was not available in PDB. Therefore, web tools such as Robetta, I-TASSER, and AlphaFold ²¹ were used to determine the 3D model of the protein. The SAVESv6.0 (PROCHECK) ²² server was used for model validation (Table 1).

OPEN IN VIEWER

Table 1.

Showing structure validation of models predicted through Robetta, AlphaFold, and I-TASSER.

Tools	PDB structure number	ERRAT	Ramachandran plot results
			Most favourable region, %	Allowed region, %	Disallowed region, %
Robetta	1	84.0604	90.8	8.6	0.6
	2	84.0871	87.4	10.6	2.0
	3	85.5025	88.8	9.6	1.6
	4	81.9835	86.6	12.2	1.2
	5		83.8	15	1.2
AlphaFold	1	94.6809	77.6	20.4	2.0
I-TASSER	1	75.6388	65.3	31.4	3.2
	2	69.0909	53.7	41.1	5.2
	3	73.4558	57.9	38.1	4.0
	4	69.9336	56.7	40.1	3.2
	5	80.2326	65.7	28.8	5.4

Abbreviation: PDB, Protein Data Bank.

Results

The sequence of TBK1-binding protein 1 was retrieved from the NCBI data bank; it is a 9-exonic protein encoded for 615 amino acids. The homologous sequence was searched by Blastp with all PDB files. The deep view of its homology search shows that TBK1-binding protein 1 was showing homology >80% with Homo sapiens.

Protein model building and validation

The 3D structure of the proteins was modelled by Robetta, an ab initio approach, ¹⁵ I-TASSER, and AlphaFold (Figure 1A to C).

OPEN IN VIEWER

Figure 1.

Three-dimensional (3D) structure predicted through (A) I-TASSER of targeted protein, (B) Robetta of targeted protein, and (C) AlphaFold of targeted protein.

The residues were categorized in the Ramachandran plot analysis based on their quadrangle regions. The graph’s red sections show the most permitted areas, whereas the yellow areas show permitted areas. Ramachandran plot was generated by PROCHECK for models developed using the Robetta, I-TASSER, and AlphaFold. Using Ramachandran map calculations conducted with the aid of the PROCHECK tool, the stereochemical quality of the predicted models and the quality of the protein models were assessed following the refinement process. In our studied protein, the total number of residues scattered in the most distributed area is greater than 85%, which shows the accuracy and high quality of the modelled structure. ²⁷

The target protein model predicted through I-TASSER has 3.2 residues in the disallowed region and ERRAT value of 75.6388; it was validated through Ramachandran plot using SAVESv6.0 (Figure 2A).

OPEN IN VIEWER

Figure 2.

The Ramachandran plot created by the PROCHECK server: (A) I-TASSER–predicted modelled protein, (B) Robetta-predicted modelled protein, and (C) AlphaFold-predicted modelled protein.

The model predicted through Robetta has 0.6 residues in the disallowed region and ERRAT value of 84.0604; it was also validated through Ramachandran plot using SAVESv6.0 (Figure 2B). The 3D structure of the protein predicted through AlphaFold has 2.0 residues in the disallowed region and ERRAT value of 94.6809 (Figure 2C).

SAVESv6.0 is the software we have used to evaluate the ERRAT value and Ramachandran plot of all the models. The model with the best Ramachandran plot was the model of Robetta.

Discussion

Proteins are important components of life and the knowledge about their structure can facilitate in the understanding about their function. Experimentally, structures of about 100 000 proteins have been determined so far. But actually, it represents a small segment of the billions of known protein sequences. The 3D structure prediction of protein depends solely on its amino acid sequences. ²⁸ Understanding a protein’s structure enables researchers to plan site-directed alterations with the goal of altering function.

In computational biology, the information is reproduced efficiently because of the available databases. Efficient bioinformatics tools and methodologies were also developed to study the 3D structures of proteins and study their functions, which was an applied approach. Our study focused on the various bioinformatics tools to design a comparative structure of TBK1-binding protein 1.

TBK1-binding protein 1 as a gene is highly expressed in mature natural killer T cells and as a protein it interacts physically with the protein kinase TBK1. ²⁹

TBK1-binding protein 1 forms complexes with different proteins such as NAP1 and TBK1 and performs various biological functions mainly involved in antiviral innate immunity, autophagy, phosphorylation, controlling viral infections, and signalling pathways. All the above functions were performed through the regulation of kinase activity. ³⁰ Three-dimensional structure of TBK1-binding protein 1 predictions through different online tools (I-TASSER, Robetta, and AlphaFold) provides insight about structural and biological functions.

In this study, AlphaFold, Robetta, and I-TASSER were preferred techniques to determine the 3D structure of an undefined protein. Structures were substantiated using a tool called SAVESv6.0 when they have been ascertained. The finest models were then shortlisted on further refinement. The technology behind AlphaFold makes predictions about the protein’s underlying physical structure and assesses which predictions were realistic using internal confidence benchmarks.

The application of bioinformatics is a practical way to study proteins in detail. Briefly, the advancements help scientists to achieve understanding of important proteins. Despite all these exciting achievements by the application of bioinformatics in biotechnology, it is still a long way to achieve. Handling proteins was very challenging because of its complexity; there was a critical need for effective bioinformatics tools which were able to provide good coverage of plant proteins. For achieving this, an advanced algorithm development is important to enable data mining, comparison, and analysis. Therefore, bioinformaticians and experts with mathematical background could play an important role in bringing practical approaches and knowledge into bioinformatics, not only for the advancement in 3D structure determination, but also for further expression analysis. The achievements in bioinformatics were not only beneficial to the field of biotechnology, but will also contribute enormously to the future of humanity.

After unveiling 3D structure of novel protein, we concluded that 3D structure of protein will help us to find out its further role other than in antiviral innate immunity. In a nut shell, we apperceive that the best 3D structure of TBK1-binding protein 1 was predicted using Robetta software.