Assessing the Quality of in Silico Produced Biomolecules: The

Subscriber access provided by Iowa State University | Library

B: Biophysics; Physical Chemistry of Biological Systems and Biomolecules

Assessing the Quality of In Silico Produced Biomolecules: The Discovery of a New Conformer Rosella Cataldo, Livia Giotta, Maria Rachele Guascito, and Eleonora Alfinito J. Phys. Chem. B, Just Accepted Manuscript • DOI: 10.1021/acs.jpcb.8b11456 • Publication Date (Web): 14 Jan 2019 Downloaded from http://pubs.acs.org on January 18, 2019

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 33 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

The Journal of Physical Chemistry

Assessing the Quality of in Silico Produced Biomolecules:

The Discovery of a New Conformer R.Cataldoa, L. Giottab, M. R. Guascitob, E.Alfinitoc,* aDepartment

of Mathematics and Physics,”Ennio De Giorgi”, University of Salento, Via

Monteroni, Lecce, Italy, I-73100. bDepartment

of Biological and Environmental Sciences and Technologies, University of

Salento,Via Monteroni, Lecce, Italy, I-73100. cDepartment

of Innovation Engineering, University of Salento, Via Monteroni, Lecce, Italy,

I-73100. *Corresponding Author: [email protected]

1 ACS Paragon Plus Environment

The Journal of Physical Chemistry 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 2 of 33

Abstract The computational procedures for predicting the 3D structure of aptamers interacting with different biological molecules have gained increasing attention in recent years. The information acquired through these methods represents a crucial input for research, especially when relevant crystallographic data are not available. A number of softwares able to perform macromolecular docking are currently accessible, leading to the prediction of the quaternary structure of complexes formed by two or more interacting biological macromolecules. Nevertheless, the scoring protocols employed for ranking the candidate structures do not always produce satisfactory results, making difficult the identification of structures that are most likely to occur in nature. In this paper, we propose a novel procedure to improve the predictive performances of computational scoring protocols, using a maximum likelihood estimate based on topological and electrical properties of interacting biomolecules. The reliability of the new computational approach, enabling the ranking of aptamer-protein configurations produced by an open source docking program, has been assessed by its successful application to a set of anti-angiopoietin aptamers, for which experimental data highlighting the sequence-dependent affinity towards the target protein are available. The procedure led to the identification of two main types of aptamer conformers involved in angiopoietin binding. Interestingly, one of these reproduces the arrangement of angiopoietin with its natural target, tyrosine kinase, while the other one is completely unexpected.The possible scenarios related to these results have been discussed.The methodology here described can be used to refine the outcomes of different computational procedures and can be applied to a wide range of biological molecules, thus representing a new tool for guiding the design of bio-inspired sensors with enhanced selectivity. 2 ACS Paragon Plus Environment

Page 3 of 33 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Introduction The development of new methods in medicine and pharmacology involves the use of very specific and targeted macromolecules, useful in diagnosis and therapy. In this framework, small fragments of RNA and DNA, able to perfectly fit a given target, and therefore called aptamers, are arousing great interest. Since 1990, the SELEX technique1 is used to generate in vitro high affinity aptamers able to bind specific biomolecules. SELEX is potentially able to produce, from an initial very large pool of oligomers, those with the highest affinity for the specific target; however, it is labor-intensive, time-consuming and requires expensive starting substances.2 This technology is an iterative process consisting of three main steps, namely selection, partitioning and amplification 3[and references therein].

An optimized SELEX technique (Cell-SELEX), leading to selected structures able to

bind a wider range of targets, has been proposed and currently used to recognize successfully the molecular signatures of diverse cell types, including cancer cells. 4-5 In competition with the SELEX-like techniques, several technological and informatics tools have been recently developed to perform a more effective selection of high affinity aptamers. These include the Quantitative Parallel Aptamer Selection System6 that allows the simultaneous analysis of thousands of aptamers, integrating two different and original techniques: a microfluidic selection and an innovative cloning sequencing. This method appears more effective than the SELEX procedure in producing high affinity aptamers. As a novel strategy to arrive to the best aptamer sequence in a single round, particular attention should be paid to the Aptamer Affinity Maturation by Resampling approach7 that combines informatics and experimental tools. Specifically, a dedicated software called Resample has been written for generating aptamer libraries, starting from a given aptamer family motif. Each library is synthesized onto a DNA microarray and tested in parallel, by using a fluorescent targeting, with a strong reduction of time and cost. 3 ACS Paragon Plus Environment


Page 4 of 33

Despite the large amount of papers dealing with the selection of new aptamers for specific biological targets, the knowledge of the physico-chemical mechanisms responsible for their binding specificity is limited. 8 A major weakness of the research concerning aptamers is the scarcity of the information available about their 3D structure, such as that provided by crystallographic data8,9. As regards the binding with the specific target, the issues that still remain open include: the role of the conformational change10, the mechanism of receptor-target selection (population-shift 11 or induced fit 12), and the role played by the timescale of conformational transitions in controlling binding mechanisms. 13 Recent molecular dynamics simulations14 pointed

out that the intrinsic flexibility of aptamers may be essential for partner recognition.

The experimental methods currently employed for measuring aptamer-protein binding equilibria have been recently reviewed.15 In parallel with the advances in the in vitro procedures for selection and evaluation of aptamer sequences, many in silico approaches have been developed to predict the 3D structure of high-affinity aptamers. 16-24 In silico predictions are of great help in the identification of aptamers with high affinity for the specific target, reducing time and cost in the screening of the aptamer families. 23 However, despite the advancements made in the last few years 24 [and references therein], lack of information still persists and prevents the full understanding of the chemical recognition mechanism. Among the computational approaches, the docking procedures are able to predict the ligand conformation, when the binding pocket of the target is given, and to score the results based on the expected biological activity.16 However, scoring still remains a quite complex task, due to a plethora of factors that affect the formation of the real ligand-target complex, ranging from the conformational change to the role of the solvent. 25


Page 5 of 33 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Furthermore, the details of the reaction route leading to the formation of the complex have not yet been clarified and a debate is still open on the two above mentioned scenarios, which at present are both considered feasible: the population-shift, also known as conformational selection, and the induced fit. The former mechanism provides for the selection of the binding-competent structure, due to its specific conformation (the conformational change occurs before the binding) whereas the latter, explores the subsequent adaptation of a loosely bound conformer to the active site of the target (the conformational change occurs after the binding). Most of the in silico procedures implement a rigid docking 21, thus depicting a population shift scenario, on the other side, different methods incorporating the induced fit mechanism have been reported.17,18 The present paper proposes a novel strategy for ranking the numerous configurations produced in silico through structure prediction procedures. In particular, it focuses on a set of four anti-Angiopoietin-2 (Ang2) aptamers and one anti-Angiopoietin-1 (Ang1) aptamer, used as a control, previously investigated by Hu and co-workers. 22 Angiopoietins are a family of four different proteins, among them, Ang1 and Ang2 have a prominent role in vasculature upregulation in many types of tumors, thus making these proteins of high interest in cancer therapy. 26 At present, the Ang1- Ang2 interaction is not completely clarified and its monitoring is of high relevance, especially for avoiding sepsis shocks in patients treated with chemotherapy. 27 Aptamers have been selected as possible inhibitors of angiopoietins, showing high affinity and selectivity in targeting specific proteins of this family. 28 Hu et al. 22 selected and analyzed three RNA mutant sequences, as well as one anti-Ang2 sequence and one anti-Ang1 sequence. By using the ZDOCK program, they 22 ranked the interaction of each sequence with Ang2 . Then, to assess the prediction accuracy, they performed binding constant measurements by means of a surface 5 ACS Paragon Plus Environment


Page 6 of 33

plasmon resonance (SPR) biosensor. 22 Conflicting results emerged from the comparison between experimental and computational data. In particular, the sequence with the highest computational score among the five tested (Sequence 2_12_35) was one of the least performing in experiments. 22 In a recent paper 29, a computational strategy is proposed, based on free software, i.e. SimRNA and AutoDock-Vina 20,21, to produce a set of possible configurations for the five aptamers investigated in. 22 This strategy returns a large number of structures for a single aptamer sequence, whose calculated mean affinity is in line with the measured one. 22 It is to be highlighted that the ranking of these structures is obtained by minimizing a single energy parameter and this could produce false-positive .30 In the present paper, we propose a novel strategy for selecting among a wide set of configurations, produced with the above mentioned or similar procedures, those having the highest probability to be found in vitro. In doing so, we will follow the trail of Proteotronics 31,

a theoretical/computational analysis of proteins and aptamers, performed by using a

complex network approach . In this framework, structure and function of biomolecules can be described simultaneously, by using a graph whose topology pictures the biomolecule backbone arrangement. The network represents a set of selected interactions between the biomolecule moieties, which may depend on the chemical environment, specifically on the presence of appropriate targets.31-38 Some relevant topological features of the configurations are used to device a novel indicator, that complements the information coming from the in silico structure prediction. We start from the analysis performed in ,29,and

22

in particular from the ranking based on the indicator named effective affinity, then,

we outline a procedure to select the configurations with the highest probability to be found in vitro. We show that a simple statistical investigation can be an appropriate support in highlighting some general traits of the configurations in silico produced, and thus in the 6 ACS Paragon Plus Environment

Page 7 of 33 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


ranking procedure. Furthermore, we stress that an optimal conformational agreement between aptamer and protein has to receive a major role in evaluating the ranking of in silico produced complexes.

Methods Materials The method is applied to the five different RNA-aptamers described in 22 and listed in Table 1. Aiming at scoring the affinity for Ang2 protein, the positive control is represented by Sequence 1, an anti-Ang2 aptamer, while the Sequence 16, specific for Ang1, is our negative control.

Table 1. The primary structure of the five studied sequences 22. The first four sequences are Ang2specific aptamers, while Sequence 16 is an Ang1-specific aptamer, used as a negative control.

Name

Sequence

Sequence 1

AAAAAACUAGCCUCAUCAGCUCAUGUGCCCCUCCGCCUGGAUCAC

Sequence 15_15_38

AAAAAGAGGACGAUGCGGAUUAGCCUCAUCAGCUCAUGUGCCGCUC

Sequence 15_12_35

AAAAAGAGGACGAUGCCGACUAGCCUCAUCAGCUCAUGUCCCCCUC

Sequence 2_12_35

AAAAAUUAACCAUCAGAUCAUGGCCCCUGCCCUCUCAAGCACCAC

Sequence 16

AAAAAACUCGAACAUUUCCACUAACCAACCAUACUAAAGCACCGC

The 3D structure of the target protein 37 is reported in the public databank 38, as a fragment of the complete Ang2 protein (from E280 to D495, chain A, entry 1Z3S), representing the receptor binding domain. Different computational tools for the 3D structure prediction are available.16-22,29 Here we refer to the procedure detailed in 29 , which proceeds in two steps: i) the 3D structures of 7 ACS Paragon Plus Environment


Page 8 of 33

the aptamers were sampled by using the SimRNA tool, a stochastic procedure based on the Replica Exchange Monte Carlo simulations

21;

ii) the lowest energy configurations are

selected from the huge number of possible structures, and rigidly docked to the receptor, by means of the AutoDock-Vina tool. 20 Both SimRNA and AutoDock-Vina are free packages. This procedure provides a huge number of possible configurations for each docked aptamer, which were reduced to a few tens by means of a drastic screening based on physico-chemical parameters. Finally the configurations were ranked based on an indicator called Effective Affinity 29 (EA), which is rooted in the free energy of the complex. The final number of configurations analyzed for each sequence is reported in Table 2.

Table 2. For each sequence reported in Table 1 , number of realizations analyzed in this paper, obtained as described in 29 , after a sharp physico-chemical selection.

Sequence

1

15_15_38

15_12_35

2_12_35

16

# configurations

20

11

21

38

27

The Proteotronics approach The Proteotronics approach is a theoretical procedure able to analyse the physical response of biomaterials in electronic devices.29,31,37 Proteotronics models the single macromolecule at the microscopic level, combining information from both structure and function. The macroscopic physical features emerge from the microscopic interactions and reflect the 3D structure. Different procedures are available in the literature 31[and references therein] , whose level of refinement ranges from chemical moieties 8 ACS Paragon Plus Environment

Page 9 of 33 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


to single atoms. Proteotronics works at the single amino-acid (nucleobase) level, which is sufficient to keep most of information useful for technological applications, with the advantage of small computational time. The procedure consists of three steps: 

Drawing the graph analogue;



Assembling the interaction network;



Solving the network.

The graph analogue 33-36 is drawn starting from the 3D structure of the biomolecule, as given by experimental or computational techniques. We select the carbon atoms C1 (Cα) to represent the position of each nucleobase (amino acid). Then, the neighbor nodes are connected with a link and the adjacency matrix can be written down.32 The neighboring statement depends on the value of a free parameter, the cut-off radius RC; two nodes are defined neighbors when closer than RC. In such a way, we have the possibility to explore different possible graphs, each of them corresponding to the same macromolecule, at different levels of activation. 35,36 The graph is then converted into an interaction network, in the present case, an electrical network to represent the flowing of charges inside the macromolecule. A resistance is assigned to each link. In particular, the resistance between a couple of nodes, say a,b, is calculated like that of a cylindrical structure of length lab, the distance between the nodes, and surface Aab, the intersection area of the spheres of radius RC, drawn around the nodes.31 About the network solution, the physical response is calculated by assuming a couple of ideal electrodes connecting the network to an external bias. The network is then solved, 9 ACS Paragon Plus Environment


Page 10 of 33

for an assigned value of RC, by using the standard Kirchhoff’s laws. The output is the network resistance, calculated in the linear regime, by using appropriate resistivity values, as detailed in.33 The biomolecule resistance depends not only on the number of links, but also on the kind of pathways the charge follows, i.e. its diffusion conveys the presence of bottlenecks, dead ends, and so on. The monitoring of the resistance was shown as an effective method to explore the global topological properties of the structure

29,33,37,

as well

as the spreading of an epidemic on a complex network gives information on the network itself.40

Results In the following we compare the ranking performed by the in silico procedure with the specific topological properties of each configuration. Link number vs Effective Affinity In the present approach, both protein and aptamer are represented by a single network; when they join to form the complex. The resulting network presents a number of links larger than the sum of the links of the respective networks, due to the contacting surface.29,33 A larger contacting surface corresponds indeed to a larger number of links, i.e. a large increase of the link number following the formation of the complex reveals a good structure complementarity between the receptor and the ligand

29

and can be

interpreted as a signal of high binding affinity. In Figure 1, EA is reported vs the relative link number, i.e. the difference between the link number of the ligand-receptor complex and the link number of the aptamer alone. Each point corresponds to a single configuration. The correlation between EA and the number of links (see also the Spearman correlation in Table 3) demonstrates that the indicator EA is effective in screening the mutual proximity between the partner biomolecules. 10 ACS Paragon Plus Environment

Page 11 of 33 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Resistance response Another kind of investigation concerns the resistance of the electrical network corresponding to a given structure (the target-free aptamer and the aptamer docked with the protein). The global resistance of the network is calculated for different values of RC, to obtain a resistance spectrum. It reflects the global network connectivity and the space distribution of links. 33,34,37 For very large values of RC, the resistance tends to an asymptotic, sequence-specific value. Figure 2 reports the ratio between the resistance of the complex and the resistance of the target-free aptamer (rcomp/rapt). Each curve represents the average over all the configurations of each sequence. It is clear that the transition to the protein-bound state results in the decrease of the resistance of the free aptamer. This trait is common to other kinds of complexes34 and says that the protein well fits the aptamer, effectively completing its network with many parallel-resistance links.



Page 12 of 33

Figure 1. The relative link number vs the effective affinity (EA), realizations . Each circle corresponds to a single configuration belonging to the five analyzed sequences. A monotonic trend is quite evident and confirmed by a positive test of the Spearman correlation (Table 3).


Page 13 of 33 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Table 3. Rank and significance of the Spearman correlation between EA and the relative link number. Calculation are performed for each sequence reported in Table1. Sequence

Rank

Significance

1

-0.50

2.93 10-2

15_15_38

-0.58

6.58 10-2

15_12_35

-0.83

2.06 10-4

2_12_35

-0.57

4.75 10-4

16

-0.48

1.49 10-2

Figure 2 shows the mean value of all the realizations of the five studied aptamers. Sequence 16, the Ang1-specific aptamer, has the smallest relative resistance in the whole RC range. The relative resistances of the Ang2-specific sequences are quite close each other, all over the spectrum. Among them, Sequence 2_12_35 has the lowest curve. On a wide Rc range (until about 50 Å), we observe that the relative resistances of all the sequences are sorted in ascending order of the binding constant Ka given in 22, and that the smaller RC, the more pronounced the differences among the structures. As a matter of fact, at small RC values, only the links necessary to obtain a stable configuration are present, while increasing RC may results in the appearing of many unnecessary links.41 In general , the possibility to discriminate between Ang1 and Ang2specific aptamers persists also in the asymptotic limit. From the above, in agreement with previous results27, 28, we conclude that resistance is an interesting tool for affinity investigation, especially in the linear regime. 13 ACS Paragon Plus Environment


Page 14 of 33

Figure 2. The relative resistance spectra of the five sequences reported in Table 1. Each curve represents the mean value of the ratio of the resistance of the protein-aptamer complex to the freetarget aptamer, Calculations are performed over all the realizations of each sequence.

Conformers We compare the features of the in silico produced structures with those of the Ang2-Tie2 complex. The full-length Angiopoietin-2 protein is composed of a single amino acid chain folded in three main domains42: the N-terminal super-clustering domain (SCD), a central coiled-coil domain (CCD), responsible for Ang2 dimerization, and a Cterminal ﬁbrinogen-related domain, which represents the receptor binding domain (RBD). Three main subdomains (A, B and P) can be identified in RBD, being P the outermost one. Crystallographic data show that the Ang2 RBD is bound to the Tie2 Ig2 domain at the level of the P-subdomain, where the Lys 469, Lys 473 and Tyr476 residues, essential for binding 38, are located. It is believed that in the Ang2 dimeric or multimeric forms, the P subdomains are oriented away from the dimerization interface, allowing the simultaneous interaction with two distinct Tie2 receptors. 38,42-43 14 ACS Paragon Plus Environment

Page 15 of 33 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


The computational procedure for the selection of aptamer conformers in the Ang2-bound 22state

has been carried out using the available 3D structure of the Ang2 RBD domain,

whose surface has been considered fully accessible for binding. We expect that in silico generated complexes reproduce at least partially the kind of binding established between Ang2 and its natural receptor. It has been shown that the Ang2 P subdomain binds at the tip of the Tie2 Ig2 arrowhead through a lock-and-key mechanism, where two complementary surfaces interact with each other with no domain rearrangements and little conformational change in both counterparts.43

Figure 3. Cartoon of the Ang2-Tie2 complex (left), with Tie2 in blue, and the corresponding contact map for RC=20Å (right). In the contact map, links internal to Ang2 fragment (from E280 to D495) are drawn in red; links internal to Tie2 fragment (from A23 to P445) are drawn in cyan. Links between the two proteins are drawn in black.

In general, we find that our aptamer-Ang2 complexes mainly exhibit two types of bindingcompetent conformers: the aptamer located on the head of the P subdomain, hereafter called hair, or the aptamer embracing the protein, thus interacting also with A and B subdomains, hereafter referred to as belt. Although only the hair conformer seems to reproduce the Ang2-Tie complex (see Figure 3), a marked difference in the EA values of 15 ACS Paragon Plus Environment


Page 16 of 33

hair and belt conformers has not been observed. A belt configuration may be characterized by a high value of EA as well as the hair configuration. As discussed above, high EA values occur when the aptamer is very close to the protein, i.e. it contains many links (see Figure 1). In Figure 3 the cartoon of the Tie2-Ang2 complex is reported with the corresponding contact map, drawn for RC=20Å. The Figure 4 presents a slide show of the most representative complexes among the 5 studied sequences and the corresponding contact maps (see also the Supporting Information). In each contact map, we have selected three regions of possible binding with decreasing level of connectivity: the first (in red) from N467 to Y482, the second (in blue), from I434 to S453, the third (in green), from V370 to G420. These regions belong to the P subdomain according to the features of the natural Ang2-receptor complex. Accordingly, a hair conformer presents most of points (i.e. network links) in these three regions; on the other side, belt configurations show preferential attachment away from the P subdomain.

Figure 4. Slide show of the most representative complexes relevant to the 5 studied sequences (top line) and of the corresponding contact maps (bottom line) at RC=20Å. From left to right: Sequence 1(8), Sequence 15_15_38 (9), Sequence 15_12_35 (4), Sequence 2_12_35 (36) and Sequence 16(3). The first 40 (41 for Sequences 15_*_*) nodes of each map compete to the aptamer, following nodes compete to the protein fragment, from E280 to D495. The boxes correspond to the three regions of binding in the Tie2-Ang2 complex (see text). 16 ACS Paragon Plus Environment

Page 17 of 33 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Scoring To go deeper inside the significance of the belt conformer, in the light of its specific features, new quality indicators can be proposed. From Figure 1, it is evident that the energy landscape explored by the procedure described in 29 is quite frustrated, with the presence of few outliers (see Figure 5). This suggests that, in addition to energy/effective affinity, further information is necessary to rank the configurations, for example, data relative to their topology. We expect that the computational procedure converges in structure and function (best 3D structure and best chemical affinity) toward the real topology of the complex. Therefore, we postulate that the configuration that best describes the actual complex would optimize not a single but a set of different descriptors, especially those that incorporate information coming from both local and global properties. Focusing on the topological properties, we look for indicators concerning both these aspects, therefore, for each configuration, we consider both the link number of each node (local information) and the network link distribution (global information). Thus, to estimate the configurations that have the highest probability to be found as real products, the mean value of these indicators, calculated as reported in the Appendix and called M, is used, together with the effective affinity EA. It should be highlighted that there is not a unique way to select the relevant topological features of the aptamer-protein complex, and other choices could be considered. However, the choice here proposed is simple and effective, as shown in the following discussion. The boxplots in Figure 5 schematically resume the distribution of M and EA for all the sequences. The values of EA were distributed between 1 and Nseq as in Eq. (A3), and called EAord, EAord=1 indicates the best score, EAord=Nseq, the worst. The same normalization was adopted for M. Outliers of M are present only in the right side of the plot, above the fourth quartile, while those of EAord can be found principally in the left side of the 17 ACS Paragon Plus Environment


Page 18 of 33

plot, below the first quartile. The configurations with the lowest value of EAord , which are also outliers, are mainly in the family of belt conformers.

Figure 5. Boxplots of the M and EAord indicators (see text). Both quantities were distributed between 0 and 1. The outliers are described by dots. The type of conformer (belt/hair) of the EA smallest values which are outliers are reported on the left. The indicators M and EAord of each configuration were reported on single bar-graph, hereafter called shadow plots (SP), because EAord is drawn with the opposite sign, like the shadow of the topological indicator. This kind of representation makes easier the configuration ranking, as explained in the following.


Page 19 of 33 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Figure 6. Slide show of the shadow plots of the considered configurations for the five sequences . Dashed lines correspond to medians, outliers are coloured in red.

In many cases, the lowest EAord value is an outlier (bottom outlier) in the Tukey definition 43

(more the 1.5 times the interquartile range (IQR), or approximately 3 standard deviations

in a Gaussian distribution), and this makes the reliability of the linked configurations doubtful. In general, we pay careful attention to the outliers, because they can signal skewed distributions rather than a statistical error in a unimodal and symmetric distribution 45.

However, in the present case, the computational strategy used for producing the

configurations does not open the door to this kind of scenario. Therefore, the configurations showing anomalous EAord values in Sequences 1, 2_12_35 and 16) are considered not interesting. On the other hand, there are many structures with low values of both EAord and M: they are the ideal candidates to represent the real complexes. 19 ACS Paragon Plus Environment


Page 20 of 33

As an outcome of our computational experiment, we observe that the frequency of belt/hair configurations depends on the specific sequence, which suggests a different ability in binding the target. Looking at Figure 6, it is quite simple to select the configurations with the highest values of both the indicators. In Table 4, the first two configurations with the best values of EA and M are reported, as well as the conformer type.

Table 4. Short list of the configurations ordered in decreasing value of the effective affinity(–EA). EA is given in arbitrary units, the configuration number (Conf) and the conformer type (Type) are reported: h means hair, b means belt.

Sequence

# Conf (EA)

Type # Conf (EA)

Type

1

8 (-72)

h

17 (-67)

h

15_15_38

9 (-52)

b

4 (-41)

h

15_12_35

4 (-66)

h

13 (-51)

b

2_12_35

36 (-57)

h

5 (-57)

b

16

3(-54)

b

5 (-52)

b

Discussion The discovery of the belt conformer opens several scenarios. First, we can imagine that each stable aptamer-Ang2 complex can be represented by a single conformer. Then, selecting for each structure only the pose that maximizes both M and EA, we obtain the belt conformer for the Ang1-specific aptamer, and the hair conformer for the Ang2-specific aptamer (see Table 4). Interestingly, the mutant sequences can exist both as hair and belt conformers. 20 ACS Paragon Plus Environment

Page 21 of 33 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Second, widening the perspective of possible events, we can imagine that, in the presence of the isolated RBD domain, the binding reaction product may appear in vitro in more than one conformer. This behavior would arise from the enhanced conformational flexibility and consequent adaptability of the aptamer, compared to the Tie2 receptor, thus paving the way to an unconventional (bifurcated) population-shift scenario. Therefore, we can select not only the configuration with the best value of EA and M, but also the runner-up. In fact , the short list contains two hair conformers for the Ang2-specific aptamer and two belt conformers for the Ang1-specific aptamer, while the mutant sequences have both belt and hair conformers (see Table 4). Third, an induced fit scenario involving a post-binding transition from the belt to the hair configuration can be ruled out, since the belt conformer in most cases does not appear “loosely-bound” with respect to the hair one, due to the high number of bonds, i.e. high value of EA. Moreover, belt and hair conformers are structurally too different to imagine a facile conversion from belt to hair with a sufficiently low activation barrier. Fourth, in a population shift perspective, the computational selection of only hair configurations could be considered an indication of high specificity (i.e. enhanced receptor mimicking ability) of the considered aptamer sequence, whereas the inability to select a stable hair conformer would represent an indication of low specificity. From this point of view, our combined energetic and topological approach is extremely successful in discriminating the best (Ang2-specific aptamer) from the worst (Ang1-specific aptamer) sequences able to bind Ang2 protein. When crossing our data with experimental affinity tests carried out by Hu et al

22,

some important issues must be considered: their SPR sensor was constructed by immobilizing the biotinylated aptamers on the metal surface exploiting the thiol-gold covalent binding and the biotin/streptavidin linkage strategy. This means that the 21 ACS Paragon Plus Environment


Page 22 of 33

Angiopoietin-2, flowed onto the functionalized sensor surface, was free to interact with the aptamer virtually in any orientation. However, it is noteworthy to point out that Hu and coworkers did not employ as ligand the Ang2 RBD domain, but the commercial full-length protein, which of course is expected to dimerize, through the CCD, and/or to form higher order multimers, whose formation is facilitated by the N-terminal SCD. Since it has been suggested that A and B subdomains in the RBD moiety could mediate interactions between the individual angiopoietin monomers in the multimeric assemblies

43,

they can be

considered partially or totally inaccessible for aptamer binding. In the light of these features, the SPR experiment does not allow to test the actual affinity of most belt conformers toward the RBD domain, as these conformers embrace RBD at the level of A and B subdomains. This finding nicely explains the noteworthy correspondence of our prediction results with experimental data, if some belt configurations are rejected based on supramolecular steric hindrance.

Conclusions This paper critically analyses the results of an in silico generation of 3D-structures of complexed biomolecules, thus proposing a procedure to identify those having the largest chance to represent the real product of the binding reaction. The aptamer-protein complex, in all its possible configurations, is mapped into an electrical network, which preserves its topological properties and simulates its resistance response, when contacted with an external bias. Each complex shows two main types of conformers: one named hair, which mimics the binding of the natural protein target (Tie2), the other named belt, which ties the protein also far from the original binding domain. The quality indicator given by the in silico 22 ACS Paragon Plus Environment

Page 23 of 33 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


procedure, here called effective affinity (EA) seems unable to discriminate between these two types of conformers. By using both the topological and electrical properties, a novel quality indicator is introduced, thus allowing to score together with EA the complexes of each possible configuration. The results are: a. the Ang2-specific aptamer prefers hair conformers; b. the Ang1-specific aptamer prefers belt conformers; c. mutant sequences do not show a distinct preference. In other terms, it seems that the hair conformer characterizes the high affinity complexes, while the belt conformer characterizes the low affinity complexes. Furthermore, both types can be present in a real sample and their relative amount reflects, in a typical population shift scenario, the affinity of the whole sample. Even if these findings clearly emphasize the well-known difficulty in making a reliable ranking of in silico produced structures, the introduction of a new quality indicator seems mandatory to discriminate between the two possible types of conformers, in which the complexes may appear. In addition, further elements have to be considered in performing the computational selection, such as the possibility for the target proteins to form multimers. This reduces the surface on which the aptamer may really be bound, thus making some of the in silico produced configurations not allowable. Based on these observations and in the present state of the art, alternative investigations are necessary to complement and interpret the results of computational structure modelling. The procedure here proposed can be considered as a complementary test to the ranking performed by standard computational approaches, useful, for instance, for increasing chances of success in designing high-specificity biosensors (aptasensors). Moreover, considering that the results are in good agreement with the present literature, it increases its merits as valuable in the absence of crystallographic data.



Page 24 of 33

Appendix In this appendix we detail the procedure for calculating the topological indicator M. As discussed in previous Sections, we combine information from local and global features of the available configurations. Concerning local information, for an assigned RC value (in present case, RC =20Å, we focus on the connectivity of each node, and compare it with its mean value, obtained performing the average all over the configurations. Therefore, we introduce an index, I1, which is calculated as follows: 𝐼1 = ∑𝑁 𝑖=𝑛|𝑥𝑖 −< 𝑥𝑖 >|

(A1)

where: xi= ∑𝒊 |

(A2)

These three indicators are finally distributed from 1 to Nseq (i.e., the number of structures for an assigned sequence, see Table 1) by using the following transformation: 𝑌𝑘 = 1 + (𝐼𝑘 − 𝑚𝑖𝑛)/Δ

(A3)

with:  =(max-min)/(Nseq-1) and min/max the min/max value of Ik. This transformation has two advantages: it uses the same range for all the indicators of a certain sequence, and preserves the distance between couples of Y-values, differently from a simple ordering. The mean value of Y1 and Y2, M= (Y1+Y2 )/2 is also evaluated. As a criterion for selecting the structures with the highest probability to be found in vitro, we propose to assume the closeness to the average of both topological and chemical indicators. To identify the best performing topological indicator, we look for that which optimizes the correlation with Y3. Therefore, for each couple of topological indicators (Y1, Y2 or M) and Y3 we can classify the configurations in LL (both quantities below the respective medians); HH (both quantities above the respective medians); and LH (one quantity below and the other above the median). The choice of median value, instead of the average, helps to prevent any misinterpretation, due to the presence of several extreme values in the indicators. In Figure A1, we plot these indicators on single bar-graph, hereafter called shadow plot (SP), in which Y3 is drawn with the opposite sign, like the shadow of the topological indicators. The draw regards Sequence 1, as a pattern of any number of similar situation. The sets of LL, LH and HH configurations of each sequence are reported as pie-charts in Figure A2. 25 ACS Paragon Plus Environment


Page 26 of 33

Figure A1. Shadow plot of Sequence 1. Lines indicate the median values of –Y3 (green), M (black), Y1 (red), Y2 (cyan). Bars represent the Y indicators as well as M: green for –Y3, black for M, red for Y1 and cyan for Y2.

Figure A2. The pie charts of LL (blue), HH (yellow) and LH (white) realizations for the 5 selected structures. M (global index) data are compared with Y 1 (topological index) and Y2 26 ACS Paragon Plus Environment

Page 27 of 33 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


(resistance index) data. With the exception of Sequence 1, the fraction of LH realizations selected by using the M indicator is equal/smaller than the fraction selected by using Y 1 and Y2. The LL and HH configurations show a convergence between the chemical and topological indicators (Figure A1). However, some not conclusive cases are present, in which the behavior of topological and chemical indicators is opposite (LH). Looking at the cumulative indicator M, we observe that the fraction of indeterminate cases (LH) is always not larger than the sum of the LL and HH cases, i.e. M optimizes the correlation with Y3 with respect Y1 and Y2, and gives more complete information about the configuration. Therefore, we can select M as the most useful indicator for searching, among the structures with very high effective affinity, those that best reproduce the expected characteristics of the real aptamer-protein complex. In other terms, we conjecture that a very high value of EA could correspond to a too extreme, not stable, topology, therefore, looking for a stable and reliable configuration, we have to maximize both the structural and chemical features.

Supporting Information Available Short list of the configurations obtained by using EA (Table S1) and different selection criteria involving both EA an M (Table S2). Selection of the contact maps for the five analyzed sequences, corresponding to both hair and belt conformers (Figures S1-S5).



Page 28 of 33

Acknowledgements Dr. Fulvio Ciriaco, University of Bari, Italy, is cordially thanked for useful discussions on the computational procedures. This research has been partially supported by the MIUR (Ministero Italiano Università e Ricerca) program FFABR 2017.


Page 29 of 33 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


References 1. Tuerk, C.; Gold, L. Systematic evolution of ligands by exponential enrichment: RNA ligands to bacteriophage T4 DNA polymerase. Science, 1990, 249, 505-510. 2. Wu, X.; Chen, J.; Wu, M.; Zhao, J. X. Aptamers: active targeting ligands for cancer diagnosis and therapy. Theranostics 2015, 5, 322. 3. Gopinath, S. C. B. Methods developed for SELEX. Anal. Bioanal. Chem. 2007, 387, 171-182. 4. Lyu, Y.; Chen, G.; Shangguan, D.; Zhang, L., Wan, S.; Wu, Y.; Zhang, H.; Duan, L.; Liu, C.; You, M. et al. Generating cell targeting aptamers for nanotheranostics using cell-SELEX. Theranostics 2016, 6, 1440-1152. 5. Shangguan, D.; Bing, T.; Zhang, N. In Aptamers Selected by Cell-SELEX for Theranostics ;Tan W., Fang X., Eds.; Springer: Berlin, Heidelberg, 2015; pp 13-33. 6. Cho, M.; Oh, S. S.; Nie, J.; Stewart, R.; Eisenstein, M.; Chambers, J.; Marth, J.D.; Walker,F.; Thomson,J.A.; Soh, H.T. Quantitative selection and parallel characterization of aptamers. Proc. Natl. Acad. Sci. USA 2013, 110, 18460-18465. 7. Kinghorn, A. B.; Dirkzwager, R. M.; Liang, S.; Cheung, Y. W.; Fraser, L. A.; Shiu, S. C.; Tang, M.S.; Tanner, J. A. Aptamer affinity maturation by resampling and microarray selection. Anal. Chem. 2016, 88, 6981-6985. 8. Cheung, Y. W.; Kwok, J.; Law, A. W.; Watt, R. M.; Kotaka, M.;Tanner, J. A. Structural basis for discriminatory recognition of Plasmodium lactate dehydrogenase by a DNA aptamer. Proc. Natl. Acad. Sci. USA 2013 ,110, 15967-15972 9. Russo Krauss, I.; Merlino, A.; Randazzo, A.; Novellino, E.; Mazzarella, L.; Sica, F. High-resolution structures of two complexes between thrombin and thrombin-binding aptamer shed light on the role of cations in the aptamer inhibitory activity. Nucleic Acids Res. 2012, 40, 8119-8128. 10. Gelinas, A. D.; Davies, D. R.; Janjic, N. Embracing proteins: structural themes in aptamer–protein complexes. Curr. Opin. Struct. Biol. 2016, 36, 122-132. 11. Catherine, A. T.; Shishido, S. N.; Robbins-Welty, G. A.; Diegelman-Parente, A. Rational design of a structure-switching DNA aptamer for potassium ions. FEBS Open Bio. 2014, 4, 788-795. 12. Lin, P. H.; Tsai, C. W.; Wu, J. W.; Ruaan, R. C.; Chen, W. Y. Molecular dynamics simulation of the induced‐fit binding process of DNA aptamer and L‐argininamide. Biotechnol. J. 2012, 7, 1367-1375. 29 ACS Paragon Plus Environment


Page 30 of 33

13. Wlodarski, T.; Zagrovic, B. Conformational selection and induced fit mechanism underlie specificity in noncovalent interactions with ubiquitin. Proc. Natl. Acad. Sci. USA 2009, 106, 19346-19351. 14. Autiero, I.; Ruvo, M.; Improta, R.; Vitagliano, L.The intrinsic flexibility of the aptamer targeting the ribosomal protein S8 is a key factor for the molecular recognition. BBAGeneral Subjects 2018, 1862, 1006-1016. 15. Jing, M.; Bowser, M. T. Methods for measuring aptamer-protein equilibria: a review. Anal. Chim. Acta 2011, 686, 9-18. 16. Kitchen, D. B.; Decornez, H.; Furr, J. R.; Bajorath, J. Docking and scoring in virtual screening for drug discovery: methods and applications. Nat. Rev. Drug Discov. 2004, 3, 935-949. 17. Totrov, M.; Abagyan, R. Flexible ligand docking to multiple receptor conformations: a practical alternative. Curr. Opin. Struct. Biol. 2008, 18, 178-184. 18. Bottegoni, G.; Kufareva, I.; Totrov, M.; Abagyan, R. A new method for ligand docking to flexible receptors by dual alanine scanning and refinement (SCARE). J. Comput. Aided Mol. Des., 2008,22, 311-325. 19. Chushak, Y.; Stone M. O. In silico selection of RNA aptamers. Nucl. Acids Res., 2009, 37, e87. 20. 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. 2010, 31, 455-461. 21. Boniecki, M. J.; Lach, G.; Dawson, W. K.; Tomala, K.; Lukasz, P.; Soltysinski, T.; Rother, J.M.; Bujnicki, J. M. SimRNA: a coarse-grained method for RNA folding simulations and 3D structure prediction. Nucl. Acids Res. 2015,44, e63. 22. Hu, W. P.; Kumar, J. V.; Huang, C. J.; Chen, W. Y. Computational selection of RNA aptamer against angiopoietin-2 and experimental evaluation. BioMed research international 2015 , Article ID 658712, http://dx.doi.org/10.1155/2015/658712 23. Strehlitz, B.; Nikolaus, N.; Stoltenburg, R. Protein detection with aptamer biosensors. Sensors 2008, 8, 4296-4307. 24. Kinghorn, A. B.; Fraser, L. A.; Lang, S.; Shiu, S. C. C.; Tanner, J. A. Aptamer bioinformatics. Int. J. Mol. Sci. 2017, 18, 2516. 25. Pagadala, N.S.; Syed, K.; Tuszynski, J.; Software for molecular docking: A review. Biophys. Rev. 2017, 9, 91-102. 30 ACS Paragon Plus Environment

Page 31 of 33 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


26. Gerald, D.; Chintharlapalli, S.; Augustin, H. G.; Benjamin, L. E. Angiopoietin-2: an attractive target for improved antiangiogenic tumor therapy. Cancer Res. 2013, 73, 1649-1657 27. Alves, B. E.; Montalvao, S. A.; Aranha, F. J.; Siegl, T. F.; Souza, C. A.; Lorand-Metze, I.; Annichino-Bizzacchi, J. M.; De Paula, E. V. Imbalances in serum angiopoietin concentrations are early predictors of septic shock development in patients with post chemotherapy febrile neutropenia. BMC infectious diseases 2010, 10, 143. 28. White, R. R.; Shan, S.; Rusconi, C. P.; Shetty, G.; Dewhirst, M. W.; Kontos, C. D.; Sullenger, B. A. Inhibition of rat corneal angiogenesis by a nuclease-resistant RNA aptamer specific for angiopoietin-2. Proc. Natl. Acad. Sci. USA 2003 100, 5028-5033. 29. Cataldo, R.; Ciriaco, F.; Alfinito, E. A validation strategy for in silico generated aptamers, Comput. Biol. Chem. 2018, 77, 123-130. 30. Kaufmann, A.; Butcher, P.; Maden, K.; Walker, S.; Widmer, M. Using in silico fragmentation to improve routine residue screening in complex matrices. J. Am. Soc. Mass Spectrom. 2017, 28, 2705-2715. 31. Alfinito, E.; Pousset J.; Reggiani, L. Proteotronics: Development of Protein-Based Electronics, Pan Stanford Publishing Pte. Ltd: New York, 2015. 32. Cataldo, R.; Alfinito, E.; Reggiani, L. Hierarchy and assortativity as new tools for binding-affinity investigation: the case of the TBA aptamer-ligand complex. IEEE Trans. Nanobiosci. 2017,16, 896-904. 33. Alfinito, E.; Reggiani, L., Cataldo, R.; De Nunzio, G.; Giotta, L.; Guascito, M. R. Modeling the microscopic electrical properties of thrombin binding aptamer (TBA) for label-free biosensors. Nanotechnology, 2017,28, 065502. 34. Alfinito, E. ; Reggiani, L. Mechanisms responsible for the photocurrent in bacteriorhodopsin. Phys. Rev. E 2015, 91, 032702. 35. Alfinito, E.; Millithaler, J. F.; Reggiani, L.; Zine, N.; Jaffrezic-Renault, N. Human olfactory receptor 17-40 as an active part of a nanobiosensor: a microscopic investigation of its electrical properties. RSC Adv. 2011,1, 123-127. 36. Alfinito, E.; Reggiani, L. Opsin vs opsin: New materials for biotechnological applications. J. Appl. Phys. 2014,116, 064901. 37. Cataldo, R.; Leuzzi, M.; Alfinito, E. Modelling and development of electrical aptasensors: a short review. Chemosensors 2018 6, 20.



Page 32 of 33

38. Barton, W. A.; Tzvetkova, D.; Nikolov, D. B. Structure of the angiopoietin-2 receptor binding domain and identification of surfaces involved in Tie2 recognition. Structure 2005,13, 825-832. 39. Berman, H. M.; Westbrook, J.; Feng, Z.; Gilliland, G.; Bhat, T. N.; Weissig, H.; Shindyalov, I.N. ; Bourne, P. E. The protein data bank. Nucl. Acids Res. 2000,28, 235242. 40. Alfinito,E.; Beccaria, M.; Macorini, G. Critical behavior in a stochastic model of vector mediated epidemics. Sci. Rep., 2016, 6, 27202. 41. Kobilka, B. K.; Deupi, X. Conformational complexity of G-protein-coupled receptors. Trends Pharmacol. Sci., 2007 ,28, 397-406. 42. Davis, S.; Papadopoulos, N.; Aldrich, T. H.; Maisonpierre, P. C.; Huang, T.; Kovac, L.; Xu,A.; Leidich,R.; Radziejewska,E ; Rafique,A.; et al. Angiopoietins have distinct modular domains essential for receptor binding, dimerization and superclustering. Nat. Struct. Mol. Biol, 2003,10, 38. 43. Barton, W. A.; Tzvetkova-Robev, D.; Miranda, E. P.; Kolev, M. V.; Rajashankar, K. R.; Himanen, J. P.; Nikolov, D. B. Crystal structures of the Tie2 receptor ectodomain and the angiopoietin-2–Tie2 complex. Nat. Struct. Mol. Biol. 2006 ,13, 524. 44. Hoaglin, D. C.; Iglewicz, B.; Tukey, J. W. Performance of some resistant rules for outlier labeling. J. Am. Stat. Assoc. 1986 ,81, 991-999. 45. Alfinito, E.; Reggiani, L. Evidence of Gumbel distributions of conductance fluctuations in bacteriorhodopsin thin films. J. Phys.: Condens. Matter, 2013, 25. 375103.


Page 33 of 33 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Table of Contents Image

33

ACS Paragon Plus Environment

Assessing the Quality of in Silico Produced Biomolecules: The

Recommend Documents