1 Introduction

Alzheimer’s disease (AD) is a neurodegenerative disorder in which your brain undergoes an irreversible deterioration causing disruption in memory and other functions and finally resulting in death from complete brain failure. The 2015 AD facts and figures suggest this as the only cause of death in the top ten in America that can neither be prevented, cured or slowed down. An alarming figure of one in every three senior citizens in America dies of AD or dementia and two in every three individual diagnosed with AD are females. An estimate of 700,000 people in the United States of age 65 and older will die in the year 2015 with AD (http://www.alz.org/). Although we do not know when the process of AD begins in an individual, it is understood that the problem begins a decade or more before the first evidence. New brain imaging techniques have given information about abundance of two abnormal structures: amyloid plaques and neurofibrillary tangles which are made of misfolded proteins in AD patients. (Alzheimer’s Disease—Unraveling the Mystery—www.nia.nih.gov/alzheimers). Another aspect of AD is the disruption in the connection between cells, which adversely affects the cell function leading to cell death. This disorder, otherwise known as “protein misfolding disorder”, is identified by the accumulation of protein aggregates at the intracellular or extracellular region [1]. Some of these protein aggregates, which contain large number of intermolecular hydrogen bond, sometimes get converted to a fibrillar structure commonly called as amyloid or beta amyloid—a misfolded structure and their accumulation makes it a plaque like structure [2]. This is observed in regions of the brain that are important in memory.

Beta amyloid (Aβ) is the peptide with 39–42 amino acids and is the major component of the amyloid plaques found in the brains of Alzheimer’s patients. This peptide is an outcome from the amyloid precursor protein (APP) by the proteolytic activities of β- and γ-secretase to produce Aβ [3, 4]. Review of literature suggests that in AD patients Aβ (1–42), fragment is the dominant one in Aβ species when compared to Aβ (1–40) in the formation of amyloid plaques. The in vitro studies displayed a dramatically increased inclination to form amyloid fibrils [5–7]. Numerous significant structural features of Aβ (1–42) fibrils have been dogged founding that Aβ (1–42) fibrils form a cross-β structure that contains parallel, in-register β-sheets [8–10]. Molecules of Aβ aggregate to form flexible soluble oligomers and may exist in several forms. We have enough reasons to believe that certain misfolded oligomers act as seeds and induce other Aβ molecules to take the misfolded oligomer form. This leads to a chain-like reaction similar to a prion infection. These seeds or the consequential amyloid plaques are lethal to the nerve cells. Tau protein, the other protein associated in AD, also forms such prion-like misfolded oligomers, and there are some evidences that the misfolding in tau is induced by Aβ [11, 12]. The plaques are a collection of a mass of regularly ordered fibrillar aggregates called amyloid fibers [13]. Latest research recommends that soluble oligomeric forms of the peptide may be instrumental in the development of AD [14, 15]. Recent studies suggest that this kind of protein aggregation in AD can be mainly caused by two factors: The misfolded proteins get accumulated between the neurons and cause disease. Secondly, this can happen because of aging resulting in decrease in proteasomal activity and alteration in the functional capacity of molecular chaperones [1]. Molecular chaperones are a class of cellular protein that plays an integral role on other proteins assisting them in their interaction, stabilization and acquiring functional conformations and leave the protein [16]. One of the major classes of molecular chaperones is heat shock proteins (HSP). Although the name says heat shock, HSPs are induced by cells in response to bacterial and viral infection, chemical exposure, deficiency of nutrition, necrosis in mammalian hosts, UV irradiation and oxidative stress [17]. HSPs are phylogenetically conserved proteins and are named according to their approximate molecular weight: HSP10, HSP40, HSP60, HSP70, HSP80 HSP90 etc. One of the most important functions of HSP is to bind with non-native proteins and assist them in acquiring their native structure and restrict them from misfolding and aggregation process. Investigation on the role of these proteins in neurodegenerative diseases revealed that they provide a basic level of defense against misfolding and are among the most potential suppressors of neurodegeneration in animal and cellular models [18, 19]. Several investigational findings have thrown light on the fact that overproduction of some HSPs can lead to the defense of numerous disease conditions including neurodegeneration progression and disease-related toxicity [20]. Latest studies have revealed that HSPs were found to be localized in protein aggregates along with cellular molecules, diseased proteins and ubiquitin. Conclusions were drawn that the HSPs probably attempted to revert the refolding of misfolded proteins or it attempted to revert the mutational effects of the gene, thus trying to suppress the disease condition; however, the mechanism is poorly understood [1]. Lately, several HSP modulators have testified to activate the solubilization of unhealthy protein aggregates. Therefore, HSPs are an excellent choice for therapeutic targeting and very likely to have strong prospective as therapeutic agents in suppressing or curing neurodegenerative disease like AD [1].

Among the various classes of HSPs, HSP60 under normal physiological conditions is an oligomer consisting of monomers arranged as two stacked complex heptameric ring [21]. This is localized to cytoplasm and mitochondrial region, and its major function is to prevent protein aggregation and to help proper folding of proteins [1]. The unfolded protein binds through hydrophobic interactions in the large cavity of the double-ring structure of HSP60 [22]. The HSP60 structure comprises of an apical domain, intermediate domain and equatorial domain. The ATP binds to the equatorial domain. The binder for equatorial and apical domains is the intermediate domain. The protein is in a hydrophobic state when inactive and the intermediate domain undergoes a conformational change when activated by ATP, which results in exposing the hydrophilic region. This assures reliability in protein binding. HSP10 supports HSP60 in folding by acting as a dome-like cover on the ATP active form of HSP60 resulting the central cavity to enlarge and help in protein folding [23].

Recently, Aβ-HSP60 conjugate vaccine was found to be effective in treating a mouse model of AD [24] Immunization with the conjugate resulted in the initiation of Aβ-specific antibodies resulting in a significant reduction in the amount of amyloid in animal model brain. This positive result encouraged testing in humans. Approximately 6 % of the vaccinated AD patients showed development of meningoencephalitis, and the human clinical trial of the active Aβ was stopped. Investigations to find out the reason for the development of pathogenic T cells attributed to the use of QS21, a very strong adjuvant and the full length of the Aβ. But, the positive results have encouraged putting in further efforts to refine the Aβ immunotherapy to produce safer vaccines for AD [24, 25].

In this current study, we have intensely studied the molecular dynamics of the Aβ-HSP60 conjugate up to 20 ns at variant time intervals, respectively. This can give us further insight into the sequential shifts if any and changes in the structural conformation of the conjugate. This study enlightens the Aβ-HSP60 conjugate as a potential therapeutic target.

2 Methods

2.1 Sequence and Structure Retrieval, Homology Modeling

The HSP60 sequence of Homo sapiens with UniprotKB [26] Accession number P10809 containing 573 amino acids was retrieved. More information about the protein was taken from Structural Classification of Protein database [27] and Superfamily 1.75. As there was no human HSP60 structure reported in Protein Data bank [28], homology modeling was carried out to model the structure. Keeping human HSP60 as the query sequence, BLAST [29] with Blosum62 substitution matrix against PDB database and default parameter was carried out to find the homologous template. After investigating the missing residues and resolution, a potential template with PDB id 1IOK was selected. Further, homology modeling using SWISS-MODEL server was carried out in automatic mode [30, 31]. The NMR solution structure of the Alzheimer’s disease Aß-peptide (1–42) was retrieved from RCSB PDB and used as the potential target in our study.

2.2 Structure Validation

Validation of the generated model was carried out using protein structure analysis (ProSA-web) tool [32, 33] and Structural Analysis and Verification Server (SAVES- Ramachandran Plot). The tools gave an understanding of the residual clashes [34]. Molecular graphics and analyses were performed with the UCSF Chimera package [35].

2.3 Molecular Dynamics

Molecular Dynamics simulation is a vital aspect to interpret the intrinsic molecular levels of several proteins. The approaches involved in understanding the atomistic level of the protein have been simplified by molecular dynamics. In our current study, the molecular dynamics simulation is used to study the functional and stability aspect of the modeled HSP60 protein. The modeled protein was exposed to dynamics study under solvent (water) environment using Gromacs 4.5.5 version [36–41].

The energy minimization of the protein was done using OPLS-AA/L (optimized potentials for liquid simulations—all-atom force field) (2001 amino acid dihedrals). The topology file thus created defines other molecules and provides system-level descriptions. The default force constant is used to keep atoms in place during equilibration, which is further illustrated during equilibration phase.

The grid preparation was done using SPC/E (extended simple point charge model) solvent model. The protein with added ions was centered in a cubic box of dimension specifying a solute-box distance of 1.0 nm with SPC/E water molecules. Four cl- ions were added as counter ions to neutralize the charges in the system using steepest descent minimization with 50,000 steps. The solvated, electro-neutral system generated is then confined to a periodic box. This was subjected to energy minimization for 5000 steps (10 ps) with steep integrator using steepest descent algorithm and continued till the potential energy was converged and stabilized. The equilibrium state of the system was achieved by keeping temperature and pressure constant for 100 ps (50,000 steps) using modified Berendsen Thermostat t-coupling algorithm and Parrinello-Rahman pressure coupling parameter. The system components were separately coupled to a temperature bath at 300 K with a coupling time constant of 0.5 ps. The short-range van der Waals and electrostatic cutoff was set to 1.2 nm. Long-range electrostatic interactions were calculated using the particle mesh Ewald (PME) method. Semi-isotropic pressure coupling was carried out using a Berendsen barostat, and the volume compressibility was chosen to be 4.5610 bar.

Comparative simulation study has been carried out with 20 ns to comprehend the structural orientation of the binding site as well as to reinstate the current finding in our research. The leapfrog integrator was used for integrating Newton’s equation in MD simulation. Linear constraint solver algorithm was used on protein covalent bonds to maintain constant bond lengths. The constraint algorithm LINCS (linear constraint solver): holonomic constraints were applied to the equilibrated protein to release the atomic constraints previously defined. Grid-type neighbor searching was applied, and long-range electrostatics was handled using PME (particle mesh Ewald). Finally, the equilibrated system was subjected to MD simulations at 300 K and all atoms were permitted to move freely for the purpose of studying the stability and its conformations.

2.4 MatchMaker

MatchMaker from the tool chimera was used for superimposing protein structures by first creating pairwise sequence alignments, then fitting the aligned residue pairs using one point per residue. A structure-based multiple sequence alignment was also computed after the structures have been superimposed. In our present study, the HSP60 protein, which was subjected to MD simulation at different time intervals, was compared in a sequential perspective using residue pair shifting with default parameters. The iteration by pruning long atom pairs until no pair exceeds x = 2.0 Armstrong was used to superimpose the structures. This tends to exclude sequence-aligned but conformational dissimilar regions such as flexible loops, allowing a tighter fit of the best-matching “core” regions [42].

2.5 Protein–Protein Interaction

Protein–protein interaction of HSP60- Aβ conjugate was carried out using Cluspro 2.0 server. [43–46]. Cluspro is a fully automated web-based program used for rigid protein–protein docking. Users have the choice of uploading the PDB files from local machine or directly from PDB server (http://www.rcsb.org/pdb/) database. The algorithm used in Cluspro 2.0 is based on fast Fourier transform using continuum electrostatics and geometric fit (DOT) [47] or a novel shape complementarity scoring function for protein–protein docking (Z-DOCK) [48] to perform rigid body docking. After processing the files, the docking algorithm evaluates multiple number of conformations retaining the most favorable interactions provides a choice to choose from balanced, electrostatic-favoured, hydrophobic-favoured and VdW + electrostaic-favoured interactions. The working of the Cluspro server is based on rotation of the ligand (protein) with 70,000 rotations relative to the receptor grid, and greedy clustering of the filtered 1000 interacted structures out of 70,000 rotations is based on positioning of ligand (protein) with 9 Å C-alpha RMSD radius until final 30 models were generated. The simulated HSP60 protein as receptor and amyloid beta (Aβ) (1–42) as a ligand were subjected to Cluspro 2.0 server with default docking parameters. The prediction used in this study was based on balanced interactions, which gave 30 interactions, out of which the most favorable interaction was the 0th model with lowest energy of −790.0 and 71 cluster sizes.

2.6 Peptide Designing and Validation

Literature review suggested the peptide amyloid beta (Aβ) (1–42) is highly immunogenic in humans. This was experimentally validated using mouse models carrying DRB1*1501 allele, which evoke T-cell response [24]. Protein Variability Server (PVS) was used to predict the antigenicity of the peptide amyloid beta (Aβ) (1–42). This method uses Kolaskar and Tongaonkar algorithm [49] based on occurrence of amino acid residues in experimentally determined epitopes. This generates an antigenic plot based on average antigenic propensity.

2.7 MHC Class II Binding Prediction

MHC class II proteins are expressed only on APCs (antigen presenting cells) that form epitopes recognized by T-helper cells (CD4+) and are involved in response to all antigens. Several conformations of peptides bound to MHC class II molecules that are having variable lengths bind to both N- and C-terminal ends. The study has been conducted in our research on the basis of the involvement of MHC class II molecules in binding to variable lengths of peptides in response to small fragments of antigen, which evoke a T-cell response. The MHC binding peptide prediction helps in finding a range of peptides defined by the property named MHC allelic polymorphism. The method used in our predictions was based on mathematical expression-based neural network approach in which the predictive accuracies are higher in comparison with other motif-based analysis and structure-based methods [50]. In this study, peptide designing of Aβ (1–15)—HSP60 (p458–474) was carried out using UCSF Chimera. Computational approaches were further used to study the effect of the constructed peptide and their interactions in human HLA. MHC class II Binding Peptide Prediction Server was used to check the residue binding positions [51]. This server helps in predicting the MHC class II binding regions in antigenic sequences.

3 Results and Discussion

3.1 Structure Validation

The assessment by Procheck server for the predicted structure gave 90.3 % residues in favorable region, 7.6 % in additional allowed region, 1.1 % in allowed and 1.1 % in disallowed region, which thus inveterate to be a good structure for further analysis. The native conformation was validated by PROSA, showing an overall model quality with a Z score of −11.27 (Figs. 1, 2, 3).

Fig. 1
figure 1

Representation of HSP60 structure using UCSF Chimera showing N- and C-terminal region

Full size image
Fig. 2
figure 2

Structural validation using procheck (Ramachandran plot analysis)

Full size image
Fig. 3
figure 3

Validation of HSP60 predicted structure using PROSA

Full size image

3.2 Structure and Dynamics

An MD simulation of the modeled HSP60 protein was performed for 20 ns in solvent environment to ensure adequate testing. The structural variation during the course of the simulation was observed to be 0.4–0.5 nm and is shown in the RMSD plot (Fig. 4). The structure is considered stable.

Fig. 4
figure 4

RMSD plot of HSP60 after 20 ns simulation using Gromacs 4.5

Full size image

3.3 MatchMaker

A search in BlastP for finding out conserved domains in HSP60 resulted in residues numbering 28–548 as conserved and showed GroEL like Type 1 chaperonin activity. This was further aided by MatchMaker tool using the residual-based comparison as shown in the supplementary files (S1: Fig. 1). The domain corresponding to the interaction with the beta amyloid, which was further implicated in our study, plays a significant role for the binding efficacy of the conjugate. The residues in comparison with wild type vary much significantly in 20 ns at C-terminal positions, but the functional binding domain region of our interest remains the same and conserved.

3.4 Protein–Protein Interaction Analysis: HSP60-Aß Conjugate

Hydrogen bonds play a vital role in predicting the interactions, shapes and properties of biomolecules [14, 15]. These are the non-covalent interactions between the hydrogen bond acceptors and donors, which determine the strength. The formation of intermolecular hydrogen bonds is observed between D…A and D–H…A. The mean hydrogen bonds distances in protein secondary structures are close to 3.0 Å according to Watson and Crick pairing model. According to Jeffrey [52], the strong hydrogen bonds with D…A distance are 2.2–2.5 Å, mostly covalent, 2.5–3.2 Å as moderate, mostly electrostatic and 3.2–4.0 Å as weak, electrostatic. The hydrogen bonding pattern of the conjugate (Fig. 5) shows a strong interaction of salt bridges formed between Arginine 141 and lysine 352 with glutamic acid 11 and aspartic acid 23, respectively, as shown in (Figs. 6, 7; Table 1). The resulted hydrogen bonding interaction pattern was generated using LigPlot+ [53] version 1.4.5 using DIMPLOT for protein–protein interaction.

Fig. 5
figure 5

Protein–protein interaction between HSP60 and Aß (1–42) peptide

Full size image
Fig. 6
figure 6

Electrostatic potential computed by Pymol showing hydrophobicity scale of the conjugate HLA-HSP60-Aß Complex. Red represents negative charged residues, and Blue represents positively charged residues

Full size image
Fig. 7
figure 7

Residue interaction plot of Aß (1–42) involved in hydrogen bonding with HSP60 using LigPlot+ (v1.4.5)

Full size image
Table 1 Hydrogen bonding interaction between amyloid beta (1–42) and HSP60
Full size table

3.5 Peptide Designing

Generally, the N- and C- terminal regions of proteins are usually solvent accessible and unstructured, Abs against those regions are also likely to recognize the native protein and are antigenic in nature. The literature study suggested that the N-terminal region of Aß Peptide (1–15) and C-terminal region of HSP60 (p458–474) are antigenic in nature. So, through computational approaches, the Aß-HSP60 conjugate was designed and visualized using UCSF Chimera as shown in (Fig. 8). N-terminal region of Aß peptide (1–15) and C-terminal region of HSP60 (p458–474) were conjugated using structure-editing tool in chimera where the designing of peptide with default φ (phi) and ψ (psi) angles i.e., φ (°) −57 and ψ (°) −47 of alpha helix conformation and Dunbrack [54] rotamer library were chosen (default parameters). The rotamer at each position will be evaluated as by the fewest number of clashes. The residues are added in N → C order and assigned a specified chain ID. The constructed peptide Aß-HSP60 conjugate was subjected to minimization using minimize tool in chimera. The default algorithm, i.e., steepest descent (10,000 steps) with a step size of 0.02 Å and conjugate gradient (100 steps) with step size of 0.02 Å, was applied for minimization. Further analyses were done using the minimized peptide Aß-HSP60 conjugate.

Fig. 8
figure 8

Aβ (1–15)–HSP60 (p458–474) conjugate peptide

Full size image

The secondary structure conformation of the peptide was depicted in (Table 2) which was done using SOPMA [55] secondary structure prediction server.

Table 2 Secondary structure analysis of the peptides
Full size table

3.6 Antigenic Site Prediction

The antigenic sites prediction of Aβ-HSP60 conjugate was done using Protein Variability Server. The average antigenic propensity was 0.9963, and it returned two antigenic peptides as shown in (Fig. 9). The graph plotted clearly indicates that the peptide sequence SGYEVHH (8–14) of Aβ was more antigenic than that of HSP60 (21–27) region.

Fig. 9
figure 9

Antigenic site prediction

Full size image

3.7 MHC Class II Binding Prediction

The constructed Aß-HSP60 peptide was subjected to MHC class II binding prediction and Rankpep [56] to analyze the MHC binding affinity of the peptide. On the basis of these results, the peptide of interest that is antigenic in nature can be proposed as a vaccine candidate. Table 3 shows the predicted peptides of Aß-HSP60 conjugates, which are having binding efficacy to MHC class II in which the third peptide IGIEIIKRT (20–28) shows highest binding affinity with a score of 9.69. The allele used was HLA-DRB1*1501 which was also used in experimental methods in mice models. The Rankpep server predicts the binding peptides based on position-specific scoring matrices (PSSMs). In addition, it predicts those MHC ligands whose C-terminal end is likely to be the result of proteasomal cleavage. On the basis of the function of MHC class II molecule, the Rankpep server returns a peptide VHHQNEDQK (12–20) that is most likely present in C-terminal region of Aß-HSP60 peptide.

Table 3 Ranking of peptides based on binding efficacy to MHC class II
Full size table

Each peak in the graphical view plot (Fig. 10) depicts the binding affinity of peptide in each peptide frame comes under the score distribution profile. This profile provides the magnitude of the binding affinity of each peptide that is helpful in finding the promiscuous binders which correlates to the immunogenicity of each peptide. The peaks under the dashed lines (blue color) are predicted as the promiscuous binders. The threshold profile is plotted on a scale of threshold vs. number of peptide frames that are helpful in selecting peptide binders. A peak crossing dashed line (red color) in threshold profile depicts the predicted binders plotted on Y-axis (peptide frame score), which can be reflected in antigenic sequences [50].

Fig. 10
figure 10

Aß-HSP60 conjugate binding to MHC class II. The rectangular box represents score distribution profile, and the square box represents the threshold profile

Full size image

3.8 Protein–Protein Interaction of Aß-HSP60 Peptide with HLA-DRB-1 Allele of Human

HLA-DRB1 (HLA class II histocompatibility antigen, DRB1-1) is a human allele of MHC class II that binds to peptides, which are antigenic in nature for recognition of CD4 T cells through endocytic pathway. The binding site of HLA can accommodate 10–30 residues. Aß-HSP60, thus immunogenic to human acts as an antigen, which further can be degraded through several degrading enzymes. Thus, the antigens that are presented to MHC class II molecules undergo proteolysis due to high lysosomal environment. In our current research, protein encoding the HLA-DRB1 gene has been retrieved from Protein Data Bank [PDB ID-1AQD (2.45Å)] as shown in (Fig. 11) that is having function of antigen processing and presenting [57]. Computationally constructed peptide Aß-HSP60 was subjected for protein–protein interaction with HLA-DRB1 allele of human using Cluspro 2.0 server. The default parameters were used as discussed earlier in Materials and Method section. The interaction resulted in 30 binding interactions out of which the lowest energy interaction with 794.9 and cluster member of 242 was selected. The interaction stemmed in binding of peptides predicted from MHC class II binding server and Rankpep. The residues LYS27, LYS31, GLN19, ASN16 and LEU30 interacted with HLA-DRB1 allele implicates that most of the residues are from C-terminal region of the antigenic peptide (Fig. 12), and the hydrogen bonding analysis was depicted in (Table 4) with H-bond distance and atoms involved in the interaction using Chimera H-bond analysis tool and Ligplot+.

Fig. 11
figure 11

Structure of HLA-DRB1 showing N- and C-terminal regions. (Blue represents Chain A, Green represents Chain B, Red represents Chain C)

Full size image
Fig. 12
figure 12

Interaction of Aß-HSP60 conjugate with HLA-DRB1

Full size image
Table 4 H-bond interactions between Aß-HSP60 with HLA-DRB1
Full size table

4 Conclusion

This study focuses on the structural perspective of Aß (1–42) peptide, which is considered as an antigenic to human cells. The utilization of Aß (1–42) peptide has several immunodominant epitopes that were experimentally validated using mice models with different genetic backgrounds. The HSP60 model was predicted using Swiss-Model server and was validated using Ramachandran plot in which most of the residues were in mostly favorable regions. In the line of study, Aß interaction with HSP60 concentrates on post-translational modifications and folding process, which necessarily do not occur in case of Alzheimer’s disease. Since the formation of electrostatic bonds with hydrogen bonds depicted the strong interaction of the Aß with HSP60, the mechanism underlying the concept is still a mystery. As per experimental data provided by Nemirovsky et al. tested in mice models, the concept of vaccine development against the aggregation of the peptide Aß was proposed and tested. The employment of Aß (1–42) for vaccine development computationally was implemented in our work. In structural perspective, the AD vaccine was modeled using N-terminal region of Aß (1–15) peptide conjugated with HSP60 (p458–474). To check the antigenicity of the modeled Aß-HSP60 peptide, server-based predictions were applied to identify the most antigenic sites based on average propensity. Several peptide regions were predicted as antigenic based on different algorithms. According to the literature review, several epitopes in humans were mostly in C-terminal regions and were HLA-DR dependent of MHC class II cells. HLA-DRB1 (HLA class II histocompatibility antigen, DRB-1) is a human allele of MHC class II that binds to peptides, which are antigenic in nature for recognition of CD4 T cells through endocytic pathway. To justify the concept of antigenicity and dependency on HLA-DB in human cells, protein–protein interactions was carried out between HLA-DRB1 and Aß-HSP60 peptide which is responsible for evoking T-cell responses in human cells through endocytic pathway. The human allele HLA-DRB1-1501 was used in MHC binding peptide server, which resulted in prediction of C-terminal regions of the Aß-HSP60 peptide. This was achieved by protein–protein interaction of HLA-DRB1 with Aß-HSP60 peptide using Cluspro 2.0. The help of binding residues then structurally visualized the immune response evoked and their bonding patterns by Aß-HSP60 peptide. The complications in developing a vaccine for specific disease like AD cannot be achieved through in silico studies, but efforts are made to simplify the time-consuming process in clinical trials through structural and sequential perspectives. Although the difficulties lie in treating AD, immunotherapy may be considered as a key factor. An ever-growing effort to apply computational methodologies having both biological and diagnostics knowledge may diminish the clinical complications in a viewpoint of molecular aspects.