An Updated Test of AMBER Force Fields and ... - ACS Publications

Jan 19, 2016 - performed to test the ability of six AMBER force fields and three implicit ...... (16) Lei, H.; Wu, C.; Liu, H.; Duan, Y. Folding Free-...
0 downloads 0 Views 7MB Size
Article pubs.acs.org/JCTC

An Updated Test of AMBER Force Fields and Implicit Solvent Models in Predicting the Secondary Structure of Helical, β‑Hairpin, and Intrinsically Disordered Peptides Irene Maffucci and Alessandro Contini* Dipartimento di Scienze Farmaceutiche − Sezione di Chimica Generale e Organica “Alessandro Marchesini”, Università degli Studi di Milano, Via Venezian, 21 20133 Milano, Italy S Supporting Information *

ABSTRACT: Replica exchange molecular dynamics simulations were performed to test the ability of six AMBER force fields and three implicit solvent models of predicting the native conformation of two helical peptides, three β-hairpins, and three intrinsically disordered peptides. Although a combination of the force field and implicit solvation models able to accurately predict the native structure of all the considered peptides was not identified, we found that the GB-Neck2 model seems to well compensate for some of the conformational biases showed by ff96 and ff99SB/ildn/ildn-φ. Indeed, the force fields of the ff99SB series coupled with GB-Neck2 reasonably discriminated helices from disordered peptides, while a good prediction of β-hairpin conformations was only achieved by performing two independent simulations: one with the ff96/GB-Neck2 combination and the other with GB-Neck2 coupled with any of the ff99SB/ildn/ildn-φ force fields.



current force fields have been validated focusing on α-helix and β-hairpin secondary structures, with less attention being paid to intrinsically disordered peptides (IDPs),26−28,33,48 principally performing the simulations in explicit solvent.7,27,28,35,37−39,41,42,46,49,50 Concerning this latter aspect, REMD simulation time dramatically increases in explicit solvent conditions; thus, large CPU power is needed when simulating long peptides or a large number of systems. Therefore, in drug discovery, the use of an implicit solvent model may be advantageous, as long as the accuracy in predicting secondary structures is maintained.51−54 Moreover, although challenging, the accurate modeling of disordered states of proteins and peptides is fundamental since IDPs are involved in important biological processes, such as signaling and regulation,55,56 and their conformational flexibility can be crucial in mediating PPIs.57,58 In light of these considerations, we tested the ability of some AMBER force fields, namely, ff96, 59 ff99SB, 36 ff99SBildn,30 ff99SBildn-φ,60 ff12SB,31 and ff14SB,32 and implicit solvation models, namely, GB-HCT,61 GB-OBC(II),62 and GB-Neck2,63 in reproducing the folding behavior of eight peptides by REMD simulations. Among these peptides, the QK VEGF modulator (H1)64 and Ac-Ala-Aib-Ala-Aib-AlaNHMe peptide (H2)65 are known to be helical; the Cterminus of protein G (B1, PDB code 2GB1),66 trpzip2 tryptophan zipper (B2, PDB code 1LE1),67 and N-terminus of ubiquitin (B3, PDB code 1UBQ)68 fold into β-hairpins, while Polybia-MPII (ID1),69 TRTK-12 CapZ peptide

INTRODUCTION Protein−protein interactions (PPIs) are involved in many biological processes, such as cell proliferation, growth, differentiation, signal transduction, and apoptosis,1−3 and they strongly depend on the secondary structure motifs at the protein−protein interface.4 Several efforts have been thus devoted to the design of peptidomimetics5 or peptide drugs,6 as well as to the development of computational methods for the prediction of the secondary structure of peptides7−12 or mini-proteins.13−16 Lately, we assisted with an improvement in computer hardware17−19 and with the development of enhanced sampling methods,20−22 aiming to overcome the limit represented by the long and CPU-intensive simulations needed to extensively sample the conformational space of peptides. Among the enhanced sampling methods, replica exchange molecular dynamics21 (REMD) has been widely applied for the reproduction of the experimental folding behavior of peptides of different lengths.8,9,21,23−29 However, a limit to the accuracy of REMD simulations in predicting peptide secondary structures might be represented by the choice of the molecular mechanics force field. The existing force fields have been mostly derived from quantum mechanics calculations or experiments, and recently, new force fields were obtained through refinement of old ones in order to improve their accuracy.30−32 Therefore, a plethora of force fields differing only in few parameters associated with specific torsion angles is currently available. Nevertheless, except for some studies,26,33−36 the comparison of force field accuracy in predicting peptide folding behavior has been focused on a limited number of test systems.27,28,37−45 Furthermore, with some exceptions,26,46,47 © 2016 American Chemical Society

Received: December 22, 2015 Published: January 19, 2016 714

DOI: 10.1021/acs.jctc.5b01211 J. Chem. Theory Comput. 2016, 12, 714−727

Article

Journal of Chemical Theory and Computation Table 1. Peptides Considered for the Present Study peptide

sequence

secondary structure

experimental data

H1 H2 B1 B2 B3 ID1 ID2 ID3

Ac-KLTWQELYQLKYKGI-NH2 Ac-Ala-Aib-Ala-Aib-Ala-NHMe GEWTYDDATKTFTVTE SWTWENGKWTWK QIFVKTLTGKTITLE INWLKLGKMVIDAL-NH2 TRTKIDWNKILS Ac-STSRHKKLMTKTE

helix 310-helix β-hairpin β-hairpin β-hairpin IDP IDP IDP

CD (water, 20 °C, pH 7.1)64 X-ray65 NMR (H2O/10% D2O, pH 6.3)66 NMR (H2O/8% D2O, pH 5.5)67 X-ray68 CD (water, 25 °C)69 NMR (H2O/10% D2O, pH 7.2)70 NMR (D2O, 37 °C)72

(ID2),70 and C-terminus of p53 (ID3)71,72 are IDPs (Table 1). In particular, the two latter peptides adopt an α-helical secondary structure when bound to the S100B protein, while they are IDP in the unbound state,70,71,73 thus representing an interesting test for the considered force fields and implicit solvent models. We decided to study only two helical peptides because modern force fields generally overpopulate the αregion;35 thus, we chose to stress more on β-hairpin and IDP predictions. Therefore, with this study, we aimed to (a) identify the most reliable combination of force field and GB model to reproduce a particular kind of secondary structure and (b) evaluate if a combination applicable to an unknown secondary structures exists or otherwise suggest how reliable predictions might be achieved. Indeed, both objectives are equally important. The former might be useful when simulating peptides with a known secondary structure, for instance, to evaluate the effect of a mutation. The latter is fundamental when doing blind predictions, such as when designing bioactive peptides from scratch.

In most cases (101 out of 144 simulations), the simulation convergence was reached within 50 ns (Table S1). For peptide H1 ff99SBildn/GB-Neck2, ff99SBildn-φ/GB-OBC(II) and ff14SB/GB-HCT simulations converged within 75 ns, whereas the ff96/GB-OBC(II) simulation converged within 100 ns. Concerning peptide H2, all the simulations converged within 50 ns. For peptide B1, ff96/GB-HCT, ff96/GB-Neck2, and ff99SBildn/GB-HCT simulations converged within 75 ns, and ff99SBildn/GB-Neck2, ff99SBildn-φ/GB-Neck2, ff12SB/ GB-HCT, and ff14SB/GB-Neck2 simulations converged within 100 ns, while the ff99SB/GB-OBC(II) converged within 150 ns. For peptide B2, ff96/GB-Neck2 and ff99SB/ GB-HCT simulations converged within 75 ns, ff96/GB-HCT and ff12SB/GB-HCT simulations converged within 100 ns, and the ff99SBildn/GB-HCT simulation converged within 150 ns. For peptide B3, ff96/GB-HCT, ff96/GB-Neck2, ff99SB/ GB-OBC(II), ff99SBildn/GB-OBC(II), ff99SBildn-φ/GBHCT, ff99SBildn-φ/GB-OBC(II), ff14SB/GB-OBC(II), and ff14SB/GB-Neck2 simulations converged within 75 ns, and ff99SB/GB-Neck2, ff99SBildn/GB-Neck2, and ff14SB/GBHCT simulations converged within 100 ns. For peptide ID1, ff96/GB-Neck2, ff99SB/GB-Neck2, ff99SBildn/GBHCT, and ff99SBildn-φ/GB-OBC(II) simulations converged within 75 ns, whereas ff96/GB-OBC(II), ff99SBildn-φ/GBHCT, and ff14SB/GB-OBC(II) simulations converged within 100 ns. For peptide ID2, the ff99SBildn-φ/GB-OBC(II) simulation converged within 75 ns, whereas ff99SB/GBOBC(II) and ff14SB/GB-OBC(II) simulations converged within 100 ns. Finally, for peptide ID3, ff96/GB-OBC(II) and ff12SB/GB-Neck2 simulations converged within 75 ns, and ff99SBildn/GB-OBC(II) and ff12SB/GB-OBC(II) simulations converged within 100 ns. Cluster analyses were conducted with cpptraj on the trajectory time intervals where the convergence criteria were met. One of every two frames was sampled by using the average-linkage algorithm and requesting five clusters; the pairwise mass-weighted RMSD on backbone heavy atoms was used as a metric. Secondary structure analyses were performed with cpptraj on the basis of DSSP. H-bonds were computed with VMD 1.9.178 on the trajectory time intervals where the convergence criteria were met. A donor−acceptor distance threshold of 4.0 Å and an angle cutoff of 30° were set, and only H-bonds with an occupancy ≥5% were considered. RoG of ID1−ID3 were computed on backbone heavy atoms with cpptraj. The following discussion is based on the simulations performed by starting from the extended conformation, but comparable conclusions can be drawn from simulations started from misfolded conformations (Supporting Information).



METHODS Two REMD simulations were performed on each of the selected peptides, built with the tLEaP module of AMBER 14,74 one by starting from an extended conformation (φ = ψ = ω = 180°) and the other by starting from a misfolded conformation (i.e., β-hairpin for H1 and H2, α-helix for B1−3 and ID1−3). The parameters for α-aminoisobutyric acid (Aib), also used in previous studies where experimental validation was reported,8,9 were downloaded from the RED database.75 For simulations with GB-HCT, GB-OBC(II), and GB-Neck2 (igb = 1, 5, and 8, respectively), the bondi, mbondi2, and mbondi3 sets of radii were used, respectively. The number of replicas (12 for peptide H2, 20 for the others) and the temperature ranges were selected through the TREMD server. 76 Each simulation considering all the mentioned combinations of force field and solvent models was run until convergence and evaluated as described below (Table S1, Supporting Information). The trajectories at 308.5 K of the REMD simulations on peptide H2 and at 300.37 K of all the other simulations were extracted and analyzed over steps of 25 ns. We considered a simulation converged when the root mean square displacement (RMSD) frequency profiles of the simulations starting from the extended and the misfolded conformations were superimposable and when the averaged secondary structure contents obtained by DSSPs differed less than 5% (Tables S2, S4, S6, S10, S14, S18, S20, S22, S24−S31) .77 Considering the structural instability of IDPs, for these systems, the radius of gyration (RoG) profiles were used as an additional proof of convergence (see S.I. for details). 715

DOI: 10.1021/acs.jctc.5b01211 J. Chem. Theory Comput. 2016, 12, 714−727

Article

Journal of Chemical Theory and Computation

Figure 1. Representative structure and population of the most populated cluster from the 300.37 K trajectory extracted from REMD simulations of peptide H1.



RESULTS AND DISCUSSION

populated cluster and its pop%, were obtained by simulations performed using the combinations ff96/GB-OBC(II), ff99SB/ GB-HCT, ff12SB/GB-Neck2, and ff14SB/GB-OBC(II) (Figure 1). These observations were confirmed by the total DSSP average helical content (htot%, Table S2), which was always above 40%. Although, the highest helicity (i.e., htot% > 60%) was obtained with the ff96/GB-OBC(II) combination and ff12SB or ff14SB, independent from the implicit solvent model. Indeed, helical H-bonds occupancies are higher than those observed for other combinations (Table S3). Moreover, when DSSP gave a high percentage of helical content (e.g., ff12SB/GB-OBC(II), ff12SB/GB-HCT, ff14SB/GB-HCT, and ff14SB/GB-Neck2), the representative structure of the second

Helical Peptides. Considering the helical nature of peptide H1 and the known helical bias of most modern force fields,35 it is not surprising that all the combinations used for the simulations of peptide H1 led to a helical conformation. Indeed, in all cases, the principal cluster obtained from the analysis of the REMD simulations had a helical representative structure and a population (pop%) higher than 50% (Figure 1). The only exceptions were represented by the combination ff99SBildn-φ/GB-HCT, which gave a helical population of 46.4%, and ff99SBildn-φ/ GB-Neck2, whose corresponding representative structure is only partially folded into a helix. The best results, in terms of both helicity of the representative structure of the most 716

DOI: 10.1021/acs.jctc.5b01211 J. Chem. Theory Comput. 2016, 12, 714−727

Article

Journal of Chemical Theory and Computation

Figure 2. Representative structure and population of the most populated cluster from the 308.5 K trajectory extracted from REMD simulations of peptide H2.

10% (Table S4), and the latter could not detect H-bonds for the ff96/GB-Neck2 simulation and only a weakly occupied Aib6 → Ala2 H-bond when GB-HCT or GB-OBC(II) were adopted (Table S5). On the other hand, for most of the other methods, representative structures of the principal cluster well reproduced the crystallographic structure, although with relatively low pop% (about 40−50%; Figure 2), except for those resulting from the analysis of ff12SB/GB-OBC(II) and ff14SB/GB-HCT simulations. However, both DSSP and Hbonds analyses showed that ff12SB and ff14SB with any implicit solvent model overestimate the α-helical content at the expense of the 310-helix, while this happened at a lower extent with the ff99SB series combined with GB-HCT and GB-Neck2 (Tables S4 and S5).

cluster (Figure S2) was partially folded into helix, positively affecting the average DSSP helical content. Furthermore, both cluster and DSSP analyses showed that the simulations with ff96/GB-HCT and ff96/GB-Neck2 predicted a significant amount of β-hairpin (Figure S2 and Table S2), which is not consistent with experimental findings.64 When considering peptide H2, which is shorter than H1 and contains the known helix stabilizer Aib,79−81 ff96 fails in predicting the helical secondary structure, independently from the solvent model. Indeed, the representative structures of the two most populated clusters showed in all cases a RMSD from the native-like structure (Figure S1) of 2.9 Å or more (Figure 2 and Figure S3). These results were confirmed by both DSSP and H-bond analyses. The former gave htot% < 717

DOI: 10.1021/acs.jctc.5b01211 J. Chem. Theory Comput. 2016, 12, 714−727

Article

Journal of Chemical Theory and Computation

Figure 3. Representative structure and population of the most populated cluster from the 300.37 K trajectory extracted from REMD simulations of peptide B1.

of helices. Therefore, it is not surprising that most of the REMD simulations performed on peptide B1 failed in predicting the native-like β-hairpin conformation,66 as observed from cluster (Figure 3), DSSP (Table S6), and Hbond (Table S8) analyses. In detail, DSSP showed that the best, although far from ideal, results were obtained with the ff96/GB-HCT simulation, with an antiparallel β-sheet content of 13.9% and an equivalent htot% (Table S6), while NMR studies in water showed a hairpin population of B1 at about 40%.66 Moreover, only two (e.g., Thr13 →Thr4 and Lys10 → Asp7) of the six backbone H-bonds possible in the native-like structure were detected by the H-bond analysis (Tables S6 and S7). A RMSD of 2.8 Å from the native conformation was found for

Therefore, the structure of medium-to-long natural helical peptides is well reproduced by any force field/GB model combination used here. Conversely, short helical peptides containing non-natural amino acids are well simulated by using any of the ff99SB/ildn/ildn-φ force fields coupled with GB-OBC(II) or GB-Neck2 models. Indeed, in line with previously reported results,8,9,12 the above combinations provide a significant amount of native-like conformations and at the same time well discriminate between the α- and 310-helix. β-Hairpins Peptides. Because of the mentioned force field helical bias35,36,82,83 and a questionable estimation of salt bridges by some implicit solvent models,26,43,44,84,85 the prediction of β-hairpins can be more challenging than that 718

DOI: 10.1021/acs.jctc.5b01211 J. Chem. Theory Comput. 2016, 12, 714−727

Article

Journal of Chemical Theory and Computation

Figure 4. Representative structure and population of the most populated cluster from the 300.37 K trajectory extracted from REMD simulations of peptide B2.

simulations, coupled to any of the considered GB models (Table S6), suggesting that this force field should seldom be used with implicit solvation. Apparently, our results disagree with those previously reported by Shell and co-workers,26 who found that the ff96/GB-OBC(II) combination was able to well reproduce the native structure of B1. However, their REMD simulations were run by starting from the native structure and requesting a 10 ns length for each replica, with analyses performed on the last ns. It has also been reported that continuum solvent models favor non-native B1 structures since they push the charged side chains to form salt bridges instead of being fully solvated, thus overwhelming the hydrophobic interactions needed to form the β-hairpin.26,44 In detail, the salt bridge between

the representative structure of the most populated cluster, which is the lowest found among all the simulations of this peptide (Figure 3) and lower than that reported in the literature for similar studies.26 Acceptable results were obtained by the ff96/GB-Neck2 combination, where the most populated cluster had a relevant pop% (64.9%) and a representative structure with a RMSD from the native-like peptide of 4.8 Å (Figure S4). Moreover, although DSSP analysis gave an average antiparallel β-sheet content of only 6.2%, it also did not show any other wellstructured motif (Table S6), consistent with results from Hbond analysis (Table S8). All the other force field/GB model combinations led to a wrong prediction, favoring α/310-helix or disordered conformations. The helical bias was particularly strong for ff14SB 719

DOI: 10.1021/acs.jctc.5b01211 J. Chem. Theory Comput. 2016, 12, 714−727

Article

Journal of Chemical Theory and Computation

Figure 5. Representative structure and population of the most populated cluster from the 300.37 K trajectory extracted from REMD simulations of peptide B3.

field/GB model combinations failed in predicting the native structure. Indeed, except for ff96 coupled to GB-HCT, or at minor extent GB-OBC(II), the representative structures of the principal cluster showed a RMSD from the native structure of about 5 Å or more (Figure 4). These results were confirmed by DSSP and H-bond analyses (Tables S10 and S12), which evidenced the presence of a reasonable amount of average antiparallel β-sheet content and a negligible helical content for ff96/GB-HCT or GB-OBC(II), while with GB-Neck2 only disordered conformations were found. For all the other simulations, a helical bias can still be noticed (htot% > 10%), with a maximum observed for ff14SB coupled with any GB model (Table S10). However, H-bond analysis evidenced i + 3 → i or i + 4 → i H-bonds with occupancies generally lower than 30%, showing that the preference for the helical

Lys10 and Asp6 brings the latter residue, which is near to the β-hairpin turn, in close contact to Lys10, causing the expulsion of Tyr5 and Phe12 side chains from the hydrophobic core.44 A H-bond between Asp6 and Lys10 with a significant occupancy was found in all the simulations except those with GB-Neck2 (Table S9), suggesting that this model allows a better description of ion pairing. However, a force field-dependent effect altering the salt-bridge populations, as already hypothesized,38 or introducing a conformational bias toward helices cannot be excluded. Indeed, the ff14SB/GB-Neck2 model gave no salt bridges, but still predicted a helical conformation (Tables S6 and S9). We also studied the folding behavior of peptide B2, which has been proved to be a stable monomeric β-hairpin by NMR experiments in water.67 In this case also most of the force 720

DOI: 10.1021/acs.jctc.5b01211 J. Chem. Theory Comput. 2016, 12, 714−727

Article

Journal of Chemical Theory and Computation

Figure 6. Representative structure and population of the most populated cluster from the 300.37 K trajectory extracted from REMD simulations of peptide ID1.

also gave acceptable performances. In detail, cluster analysis performed on the ff96/GB-HCT simulation resulted in a principal cluster with a rather high pop% (89.9%) and a representative structure that deviates from the native structure of only 2.0 Å (Figure 5). Similar, although slightly worse, results were obtained with ff96 and either GB-OBC(II) or GB-Neck2. DSSP analysis showed that the simulations performed with ff96 had a total average β-sheet content of about 20−30%, with the highest and lowest percentages obtained for GB-HCT and GB-OBC(II), respectively, and a low htot% (max 6%) (Table S14). Only H-bond analysis of the ff96/GB-OBC(II) trajectory evidenced the presence of five out of the six native H-bonds involving the peptide backbone, although with poor occupancies, while none of the native Hbonds was found with GB-Neck2 (Tables S15 and S16).

secondary structure, although present (Table S12), is less marked for this system compared to B1 and probably limited to a force field effect. Indeed, no persistent ionic interactions were observed in B2 simulations except those with ff12SB or ff14SB coupled to any GB model and to a minor extent the one with ff99SB/GB-OBC(II) (Table S13), where a salt bridge between Lys8 and Glu5 was sampled. However, while the two former methods predicted a high helical content, the latter ended in an acceptable native-like β-hairpin conformation. As an additional β-hairpin example, we simulated peptide B3, which is the N-terminal sequence of ubiquitin.68 Consistent to what was observed for B2 and at a minor extent B1, the best results were obtained with the ff96 force field coupled with GB-HCT, although the other GB models 721

DOI: 10.1021/acs.jctc.5b01211 J. Chem. Theory Comput. 2016, 12, 714−727

Article

Journal of Chemical Theory and Computation

Figure 7. Representative structure and population of the most populated cluster from the 300.37 K trajectory extracted from REMD simulations of peptide ID2.

force field/GB model combinations gave an average antiparallel β-sheet content of about 20%, although a certain amount of helix was at the same time found (htot% > 10%; Table S14), consistent with results from H-bond analysis (Table S16). Except for ff99SBildn-φ/GB-Neck2, which also behaved fairly, the other combinations showed a preference for the helical conformation, as evidenced by cluster, DSSP, and Hbond analyses (Figure 5 and Figure S6 and Tables S14 and S16). This might be due the combined effect of the force field biases and the salt bridge overestimation by the implicit solvent model, with ff14SB being the most helical and GBHCT the most salt bridge stabilizer. Indeed, salt bridges between Lys5 or Lys10 and Glu15 were sampled in all simulations performed with GB-HCT, except those based on

However, helical H-bonds were poorly sampled as well, while some H-bonds, principally involving Leu14 and Phe3 or Lys5 and Ile12, were identified, particularly by using ff96/GBNeck2 (Table S16). This is indicative of the presence of βhairpin-like conformations with the turn between the two βstrands being shifted toward the N-terminus, compared to the native structure (Figure S7). Contrary to what was observed for peptides B1 and B2, ff99SBildn/GB-Neck2 and ff99SB/GB-Neck2 were also able to reasonably predict a native-like conformation for peptide B3. Indeed, the principal cluster (pop% = 35.5%) obtained by the latter method and the second cluster (pop% = 41.4%) obtained by the former had a representative structure with rather low RMSDs from the native structure (2.4 and 3.7 Å, respectively; Figure 5 and Figure S6). Furthermore, these 722

DOI: 10.1021/acs.jctc.5b01211 J. Chem. Theory Comput. 2016, 12, 714−727

Article

Journal of Chemical Theory and Computation

Figure 8. Representative structure and population of the most populated cluster from the 300.37 K trajectory extracted from REMD simulations of peptide ID3.

of β-sheet content, only marginally compensated by other secondary structures (Table S18), and the radius of gyration profiles showed peaks suggesting the presence of compact structures (Figure S10). Although less intense, analogue peaks were observed for ff12SB and ff14SB, which in this case also showed a helical bias. Indeed, the representative structure of the most populated cluster (pop% > 70%) obtained from the analysis of the related trajectories (Figure 6) was helical. DSSP analysis also gave levels of helical propensity significantly higher than the average percentages of other structures (htot% > 40%; Table S18). Similar observations can be done for ff99SB and ff99SBildnφ coupled with GB-HCT or GB-OBC(II) and for ff99SBildn coupled with GB-OBC(II), although htot% was lower than that

the ff96 force field (Table S17), thus leading to misfolded conformations (Figure S8). In light of this, ff96/GB-HCT appears to be the best force field/implicit solvent model combination when simulating β-hairpins. Intrinsically Disordered Peptides. The discrimination of disordered from well-structured peptide states is also of fundamental interest because of the role of IDPs in biological events.46,47,55−58 Moreover, testing the force field/GB model combinations on IDPs allows a better evaluation of their eventual biases toward a particular secondary structure. Indeed, when considering ID1, ff96 coupled with any implicit solvent model favored a β-hairpin secondary structure. Cluster analysis always gave highly populated top clusters whose representative structures were β-hairpin-like geometries (Figure 6). Moreover, DSSP analysis showed a high amount 723

DOI: 10.1021/acs.jctc.5b01211 J. Chem. Theory Comput. 2016, 12, 714−727

Article

Journal of Chemical Theory and Computation observed for ff12SB and ff14SB force fields (Table S18), and the representative structures of the main clusters were not perfectly helical (Figure 6). Conversely, ff99SBildn coupled with GB-HCT or GB-Neck2 gave a representative structure of the most populated cluster, which was helical only at the Cterminus (Figure 6), while for ff99SBildn/GB-Neck2, a βhairpin was obtained as the representative conformation of the second cluster (Figure S9), in agreement with the average secondary structure amount obtained by DSSP (Table S18). The ff99SBildn-φ simulations gave similar results, although a geometry with a high helical content was obtained as the representative structure of the principal cluster (Figure 6). Hbond analysis found H-bonds corresponding to both turn- and β-sheet-like conformations with moderately low occupancies (Table S19), thus explaining the broader distribution of the radius of gyrations profiles (Figure S10). Therefore, although a trajectory without a detectable amount of defined secondary structures was never obtained for ID1, the ff99SBildn and ff99SBildn-φ force fields provided an acceptable description of a disordered conformation. In particular, when using the GB-Neck2 model, the trajectory analyses showed a similar amount of helical and β-hairpin content, which can be interpreted as a warning of structural instability and a suggestion that the system under study is an IDP. Conversely, ff96, ff12SB, and ff14SB combined to any GB model do not seem to be suited to simulate unstructured peptides. The analyses performed on ID2 and ID3 trajectories gave similar results, although some differences were found. First of all, the ff96/GB-OBC(II) combination unexpectedly turned out to have a helical preference, very evident for ID2 (Figure 7 and Figure S11 and Table S20) and weak for ID3 (Figure 8 and Figure S13 and Table S22). On the contrary, ff96/GB-Neck2 was the best combination in predicting the disordered conformation of both ID2 and ID3, as shown by DSSP analysis, where no preferential secondary structure was found (Tables S20 and S22), by the radius of gyration profile, which had a rather broad distribution (Figures S12 and S14), and by the absence of persistent H-bonds (Tables S21 and S23). The ff99SB, ff99SBildn and ff99SBildn-φ force fields coupled with GBNeck2 also represent an acceptable choice when simulating IDPs, although a small bias toward the helical structure was observed (Figures 7 and8 and Figures S11 and S13 and Tables S20 and S22).

native salt-bridge-trap conformations,26,44,53 thus enhancing (or not compensating) the helical propensity of the force field; on the contrary, GB-Neck2 better compensated the mild helical bias showed by the ff99SB series. Finally, the ff96’s preference for a defined secondary structure was strongly dependent on the system nature, while apparently the kind of implicit solvent model marginally affected the simulations outcome. Although we did not identify a force field/GB solvent combination equally accurate in predicting the secondary structure of helical peptides, β-hairpins, and IDP, we found that ff99SB, ff99SBildn, and ff99SBildn-φ coupled with the GB-Neck2 model is a reasonably balanced combination to predict peptide folding preferences. Indeed, even if it still presents a certain bias toward helical conformations, a careful analysis of the trajectory by clustering, DSSP, H-bond, and radius of gyration can provide enough evidence on the conformational nature of an hypothetical peptide with unknown geometry; a well-structured helix as the representative structure of the main cluster together with a relatively high main cluster population and DSSP helical content most likely indicate a true helical peptide. Conversely, if one or more of these criteria are not satisfied, it might mean that the peptide is either a β-hairpin or an IDP. If this occurs, it might be helpful to perform additional simulations using ff96 coupled with GB-OBC(II) or GB-Neck2 since those combinations are able to discriminate structured from unstructured conformations and to predict a β-hairpin when appropriate. Finally, although the constantly increasing computational power seems to shade the interest toward implicit models, we believe that those methods are still attractive, especially in the field of drug design, where faster simulations allow one to explore more chemical entities per time unit. For this reason, we suggest the development of a force field specifically designed to work with an implicit solvent model (for example, the well-performing GB-Neck2), combining the good predictive power of ff96 for β-hairpins and the ability of the ff99SB series to discriminate helices from IDPs.



ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.jctc.5b01211. Representative structures of the second clusters, DSSP analysis, H-bond analysis, radius of gyration profiles, and clusters obtained from cluster analysis. (PDF)



CONCLUSIONS In the field of drug discovery and design much attention is being paid to the development of peptide drugs;86−90 thus, the availability of fast and accurate methods to predict and reproduce the secondary structure of both natural and artificial peptides represents a fundamental goal. In light of this, we evaluated the ability of several AMBER force fields (e.g., ff96, ff99SB, ff99SBildn, ff99SBildn-φ, ff12SB, and ff14SB) combined with three implicit solvent models, (GB-HCT,61 GB-OBC(II)62 and GB-Neck263) to predict the native secondary structure of 8 peptides: 2 helices, 3 βhairpins, and 3 IDPs. We found that ff12SB and ff14SB always resulted markedly helical when combined with GB implicit solvation. Conversely, the flavor of the GB model highly affected the results of simulations run with one of the ff99SB force fields series. Indeed, GB-HCT and GB-OBC(II) showed to favor non-



AUTHOR INFORMATION

Corresponding Author

*E-mail [email protected]. Tel. +39.02.503.14480. Fax: +39.02.503.1447. Author Contributions

The manuscript was written through contributions of all authors. All authors have given approval to the final version of the manuscript. Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS This work was partially supported by Ministero dell’Università e della Ricerca (PRIN 2010 − “Synthesis and biomedical 724

DOI: 10.1021/acs.jctc.5b01211 J. Chem. Theory Comput. 2016, 12, 714−727

Article

Journal of Chemical Theory and Computation

(17) Shaw, D. E.; Dror, R. O.; Salmon, J. K.; Grossman, J. P.; Mackenzie, K M, et al. Millisecond-Scale Molecular Dynamics Simulations on Anton. In Proceedings of the Conference on High Performance Computing, Networking, Storage and Analysis (SC09); ACM: New York, 2009. (18) Stone, J. E.; Hardy, D. J.; Ufimtsev, I. S.; Schulten, K. GPUAccelerated Molecular Modeling Coming of Age. J. Mol. Graphics Modell. 2010, 29, 116−125. (19) Shaw, D. E.; Maragakis, P.; Lindorff-Larsen, K.; Piana, S.; Shan, Y.; Wriggers, W.; et al. Atomic-Level Characterization of the Structural Dynamics of Proteins. Science 2010, 330, 341−346. (20) Lei, H.; Duan, Y. Improved Sampling Methods for Molecular Simulation. Curr. Opin. Struct. Biol. 2007, 17, 187−191. (21) Sugita, Y.; Okamoto, Y. Replica-Exchange Molecular Dynamics Method for Protein Folding. Chem. Phys. Lett. 1999, 314, 141−151. (22) Prinz, J. H.; Wu, H.; Sarich, M.; Keller, B.; Senne, M.; Held, M.; Chodera, J. D.; Schutte, C.; Noé, F. Markov Models of Molecular Kinetics: Generation and Validation. J. Chem. Phys. 2011, 134, 174105. (23) Ding, F.; Tsao, D.; Nie, H.; Dokholyan, N. V. Ab Initio Folding of Proteins with All-Atom Discrete Molecular Dynamics. Structure 2008, 16, 1010−1018. (24) Lin, E.; Shell, M. S. Convergence and Heterogeneity in Peptide Folding with Replica Exchange Molecular Dynamics. J. Chem. Theory Comput. 2009, 5, 2062−2073. (25) Maffucci, I.; Pellegrino, S.; Clayden, J.; Contini, A. Mechanism of Stabilization of Helix Secondary Structure by Constrained CαTetrasubstituted A-Amino Acids. J. Phys. Chem. B 2015, 119, 1350− 1361. (26) Shell, M. S.; Ritterson, R.; Dill, K. A. A Test on Peptide Stability of AMBER Force Fields with Implicit Solvation. J. Phys. Chem. B 2008, 112, 6878−6886. (27) Yoda, T.; Sugita, Y.; Okamoto, Y. Comparisons of Force Fields for Proteins by Generalized-Ensemble Simulations. Chem. Phys. Lett. 2004, 386, 460−467. (28) Yoda, T.; Sugita, Y.; Okamoto, Y. Secondary-Structure Preferences of Force Fields for Proteins Evaluated by GeneralizedEnsemble Simulations. Chem. Phys. 2004, 307, 269−283. (29) Maffucci, I.; Clayden, J.; Contini, A. Origin of Helical Screw Sense Selectivity Induced by Chiral Constrained Cα-Tetrasubstituted A-Amino Acids in Aib-Based Peptides. J. Phys. Chem. B 2015, 119, 14003−14013. (30) Lindorff-Larsen, K.; Piana, S.; Palmo, K.; Maragakis, P.; Klepeis, J. L.; Dror, R. O.; Shaw, D. E. Improved Side-Chain Torsion Potentials for the Amber ff99SB Protein Force Field. Proteins: Struct., Funct., Genet. 2010, 78, 1950−1958. (31) Zgarbová, M.; Otyepka, M.; Šponer, J.; Mládek, A.; Banás,̌ P.; Cheatham, T. E.; Jurečka, P. Refinement of the Cornell et Al. Nucleic Acids Force Field Based on Reference Quantum Chemical Calculations of Glycosidic Torsion Profiles. J. Chem. Theory Comput. 2011, 7, 2886−2902. (32) Maier, J. A.; Martinez, C.; Kasavajhala, K.; Wickstrom, L.; Hauser, K. E.; Simmerling, C. ff14SB: Improving the Accuracy of Protein Side Chain and Backbone Parameters from ff99SB. J. Chem. Theory Comput. 2015, 11, 3696−3713. (33) Lindorff-Larsen, K.; Maragakis, P.; Piana, S.; Eastwood, M. P.; Dror, R. O.; Shaw, D. E. Systematic Validation of Protein Force Fields against Experimental Data. PLoS One 2012, 7, 1−6. (34) Lange, O. F.; van der Spoel, D.; de Groot, B. L. Scrutinizing Molecular Mechanics Force Fields on the Submicrosecond Timescale with NMR Data. Biophys. J. 2010, 99, 647−655. (35) Best, R. B.; Buchete, N.-V.; Hummer, G. Are Current Molecular Dynamics Force Fields Too Helical? Biophys. J. 2008, 95, L07−L09. (36) Hornak, V.; Abel, R.; Okur, A.; Strockbine, B.; Roitberg, A.; Simmerling, C. Comparison of Multiple Amber Force Fields and Development of Improved Protein Backbone Parameters. Proteins: Struct., Funct., Genet. 2006, 65, 712−725.

applications of tumor-targeting peptidomimetics.” Prot. 2010NRREPL) and by Università degli Studi di Milano (Piano Sviluppo − “Mitochondria targeting peptide based nanomaterials”). We acknowledge CINECA and the Regione Lombardia award, under the LISA initiative, for the availability of high performance computing resources and support.



ABBREVIATIONS REMD, replica exchange molecular dynamics; DSSP, define secondary structure of proteins; PPI, protein−protein interaction; IDP, intrinsically disordered peptide



REFERENCES

(1) Schuster-Böckler, B.; Bateman, A. Protein Interactions in Human Genetic Diseases. Genome Biol. 2008, 9, 1−12. (2) Wells, J. a; McClendon, C. L. Reaching for High-Hanging Fruit in Drug Discovery at Protein-Protein Interfaces. Nature 2007, 450, 1001−1009. (3) Ferri, N.; Corsini, A.; Bottino, P.; Clerici, F.; Contini, A. Virtual Screening Approach for the Identification of New Rac1 Inhibitors. J. Med. Chem. 2009, 52, 4087−4090. (4) Davis, J. M.; Tsou, L. K.; Hamilton, A. D. Synthetic NonPeptide Mimetics of Alpha-Helices. Chem. Soc. Rev. 2007, 36, 326− 334. (5) Ruffoni, A.; Ferri, N.; Bernini, S. K.; Ricci, C.; Corsini, A.; Maffucci, I.; Clerici, F.; Contini, A. 2-Amino-3-(phenylsulfanyl)norbornane-2-Carboxylate: An Appealing Scaffold for the Design of Rac1-Tiam1 Protein-Protein Interaction Inhibitors. J. Med. Chem. 2014, 57, 2953−2962. (6) Craik, D. J.; Fairlie, D. P.; Liras, S.; Price, D. The Future of Peptide-Based Drugs. Chem. Biol. Drug Des. 2013, 81, 136−147. (7) Simmerling, C.; Strockbine, B.; Roitberg, A. E. All-Atom Structure Prediction and Folding Simulations of a Stable Protein. J. Am. Chem. Soc. 2002, 124, 11258−11259. (8) Ruffoni, A.; Contini, A.; Soave, R.; Lo Presti, L.; Esposto, I.; Maffucci, I.; Nava, D.; Pellegrino, S.; Gelmi, M. L.; Clerici, F. Model Peptides Containing the 3-Sulfanyl-Norbornene Amino Acid, a Conformationally Constrained Cysteine Analogue Effective Inducer of 310-Helix Secondary Structures. RSC Adv. 2015, 5, 32643−32656. (9) Pellegrino, S.; Contini, A.; Clerici, F.; Gori, A.; Nava, D.; Gelmi, M. L. 1H-Azepine-4-Amino-4-Carboxylic Acid: A New A,αDisubstituted Ornithine Analogue Capable of Inducing Helix Conformations in Short Ala-Aib Pentapeptides. Chem. - Eur. J. 2012, 18, 8705−8715. (10) Nguyen, H.; Maier, J.; Huang, H.; Perrone, V.; Simmerling, C. Folding Simulations for Proteins with Diverse Topologies Are Accessible in Days with a Physics-Based Force Field and Implicit Solvent. J. Am. Chem. Soc. 2014, 136, 13959−13962. (11) Best, R. B. Atomistic Molecular Simulations of Protein Folding. Curr. Opin. Struct. Biol. 2012, 22, 52−61. (12) Pellegrino, S.; Bonetti, A.; Clerici, F.; Contini, A.; Moretto, A.; Soave, R.; Gelmi, M. L. 1 H -Azepine-2-Oxo-5-Amino-5-Carboxylic Acid: A 3 10 Helix Inducer and an Effective Tool for Functionalized Gold Nanoparticles. J. Org. Chem. 2015, 80, 5507−5516. (13) Chowdhury, S.; Lee, M. C.; Xiong, G.; Duan, Y. Ab Initio Folding Simulation of the Trp-Cage Mini-Protein Approaches NMR Resolution. J. Mol. Biol. 2003, 327, 711−717. (14) Jayachandran, G.; Vishal, V.; Pande, V. S. Using Massively Parallel Simulation and Markovian Models to Study Protein Folding: Examining the Dynamics of the Villin Headpiece. J. Chem. Phys. 2006, 124, 164902−164912. (15) Pitera, J. W.; Swope, W. Understanding Folding and Design: Replica-Exchange Simulations of “Trp-Cage” Miniproteins. Proc. Natl. Acad. Sci. U. S. A. 2003, 100, 7587−7592. (16) Lei, H.; Wu, C.; Liu, H.; Duan, Y. Folding Free-Energy Landscape of Villin Headpiece Subdomain from Molecular Dynamics Simulations. Proc. Natl. Acad. Sci. U. S. A. 2007, 104, 4925−4930. 725

DOI: 10.1021/acs.jctc.5b01211 J. Chem. Theory Comput. 2016, 12, 714−727

Article

Journal of Chemical Theory and Computation (37) Kührová, P.; De Simone, A.; Otyepka, M.; Best, R. B. ForceField Dependence of Chignolin Folding and Misfolding: Comparison with Experiment and Redesign. Biophys. J. 2012, 102, 1897−1906. (38) Lwin, T.; Luo, R. Force Field Influences in B-Hairpin Folding Simulations. Protein Sci. 2006, 15, 2642−2655. (39) Martín-García, F.; Papaleo, E.; Gomez-Puertas, P.; Boomsma, W.; Lindorff-Larsen, K. Comparing Molecular Dynamics Force Fields in the Essential Subspace. PLoS One 2015, 10, e0121114. (40) Piana, S.; Lindorff-Larsen, K.; Shaw, D. E. How Robust Are Protein Folding Simulations with Respect to Force Field Parameterization? Biophys. J. 2011, 100, L47−L49. (41) Raucci, R.; Colonna, G.; Castello, G.; Costantini, S. Peptide Folding Problem: A Molecular Dynamics Study on Polyalanines Using Different Force Fields. Int. J. Pept. Res. Ther. 2013, 19, 117− 123. (42) Todorova, N.; Legge, F. S.; Treutlein, H.; Yarovsky, I. Systematic Comparison of Empirical Forcefields for Molecular Dynamic Simulation of Insulin. J. Phys. Chem. B 2008, 112, 11137−11146. (43) Zagrovic, B.; Sorin, E. J.; Pande, V. Beta-Hairpin Folding Simulations in Atomistic Detail Using an Implicit Solvent Model. J. Mol. Biol. 2001, 313, 151−169. (44) Zhou, R.; Berne, B. J. Can a Continuum Solvent Model Reproduce the Free Energy Landscape of a B-Hairpin Folding in Water? Proc. Natl. Acad. Sci. U. S. A. 2002, 99, 12777−12782. (45) Ono, S.; Nakajima, N.; Higo, J.; Nakamura, H. Peptide FreeEnergy Profile Is Strongly Dependent on the Force Field: Comparison of C96 and AMBER95. J. Comput. Chem. 2000, 21, 748−762. (46) Henriques, J.; Cragnell, C.; Skepö, M. Molecular Dynamics Simulations of Intrinsically Disordered Proteins: Force Field Evaluation and Comparison with Experiment. J. Chem. Theory Comput. 2015, 11, 3420−3431. (47) Palazzesi, F.; Prakash, M. K.; Bonomi, M.; Barducci, A. Accuracy of Current All-Atom Force-Fields in Modeling Protein Disordered States. J. Chem. Theory Comput. 2015, 11, 2−7. (48) Best, R. B.; Mittal, J. Balance between Alpha and Beta Structures in Ab Initio Protein Folding. J. Phys. Chem. B 2010, 114, 8790−8798. (49) Cino, E. a.; Choy, W. Y.; Karttunen, M. Comparison of Secondary Structure Formation Using 10 Different Force Fields in Microsecond Molecular Dynamics Simulations. J. Chem. Theory Comput. 2012, 8, 2725−2740. (50) Yang, C.; Kim, E.; Pak, Y. Probing A/B Balances in Modified Amber Force Fields from a Molecular Dynamics Study on a Bβ A Model Protein (1FSD). Bull. Korean Chem. Soc. 2014, 35, 1713− 1719. (51) Zhou, R. Free Energy Landscape of Protein Folding in Water: Explicit vs. Implicit Solvent. Proteins: Struct., Funct., Genet. 2003, 53, 148−161. (52) Ozkan, S. B.; Wu, G. A.; Chodera, J. D.; Dill, K. a. Protein Folding by Zipping and Assembly. Proc. Natl. Acad. Sci. U. S. A. 2007, 104, 11987−11992. (53) Jang, S.; Shin, S.; Pak, Y. Molecular Dynamics Study of Peptides in Implicit Water: Ab Initio Folding of B-Hairpin, B-Sheet, and Bβα-Motif. J. Am. Chem. Soc. 2002, 124, 4976−4977. (54) Jang, S.; Kim, E.; Pak, Y. Free Energy Surfaces of Miniproteins with a Betabetaalpha Motif: Replica Exchange Molecular Dynamics Simulation with an Implicit Solvation Model. Proteins: Struct., Funct., Genet. 2006, 62, 663−671. (55) Dyson, H. J.; Wright, P. E. Intrinsically Unstructured Proteins and Their Functions. Nat. Rev. Mol. Cell Biol. 2005, 6, 197−208. (56) Tompa, P. Intrinsically Disordered Proteins: A 10-Year Recap. Trends Biochem. Sci. 2012, 37, 509−516. (57) De Simone, A.; Kitchen, C.; Kwan, A. H.; Sunde, M.; Dobson, C. M.; Frenkel, D. Intrinsic Disorder Modulates Protein SelfAssembly and Aggregation. Proc. Natl. Acad. Sci. U. S. A. 2012, 109, 6951−6956.

(58) Barducci, A.; Bonomi, M.; Prakash, M. K.; Parrinello, M. FreeEnergy Landscape of Protein Oligomerization from Atomistic Simulations. Proc. Natl. Acad. Sci. U. S. A. 2013, 110, E4708−E4713. (59) Kollman, P. A. Advances and Continuing Challenges in Achieving Realistic and Predictive Simulations of the Properties of Organic and Biological Molecules. Acc. Chem. Res. 1996, 29, 461− 469. (60) Nerenberg, P. S.; Head-gordon, T. Optimizing Protein Solvent Force Fields to Reproduce Intrinsic Conformational Preferences of Model Peptides. J. Chem. Theory Comput. 2011, 7, 1220−1230. (61) Hawkins, G. D.; Cramer, C. J.; Truhlar, D. G. Pairwise Solute Descreening of Solute Charges from a Dielectric Medium. Chem. Phys. Lett. 1995, 246, 122−129. (62) Onufriev, A.; Bashford, D.; Case, D. A. Exploring Protein Native States and Large-Scale Conformational Changes with a Modified Generalized Born Model. Proteins: Struct., Funct., Genet. 2004, 55, 383−394. (63) Nguyen, H.; Roe, D. R.; Simmerling, C. Improved Generalized Born Solvent Model Parameters for Protein Simulations. J. Chem. Theory Comput. 2013, 9, 2020−2034. (64) D’Andrea, L. D.; Iaccarino, G.; Fattorusso, R.; Sorriento, D.; Carannante, C.; Capasso, D.; Trimarco, B.; Pedone, C. Targeting Angiogenesis: Structural Characterization and Biological Properties of a de Novo Engineered VEGF Mimicking Peptide. Proc. Natl. Acad. Sci. U. S. A. 2005, 102, 14215−14220. (65) Francis, A. K.; Iqbal, M.; Balaram, P.; Vijayan, M. Crystal Structure of Boc-Ala-Aib-Ala-Aib-Aib-Methyl Ester, a Pentapeptide Fragment of the Channel-Forming Ionophore Suzukacillin. Biopolymers 1983, 22, 1499−1505. (66) Blanco, F. J.; Rivas, G.; Serrano, L. A Short Linear Peptide That Folds into a Native Stable Beta-Hairpin in Aqueous Solution. Nat. Struct. Biol. 1994, 1, 584−590. (67) Cochran, A. G.; Skelton, N. J.; Starovasnik, M. A. Tryptophan Zippers: Stable, Monomeric B-Hairpins. Proc. Natl. Acad. Sci. U. S. A. 2001, 98, 5578−5583. (68) Vijay-Kumar, S.; Bugg, C. E.; Cook, W. J. Structure of Ubiquitin Refined at 1.8 Å Resolution. J. Mol. Biol. 1987, 194, 531− 544. (69) De Souza, B. M.; Da Silva, A. V. R.; Resende, V. M. F.; Arcuri, H. A.; dos Santos Cabrera, M. P.; Ruggiero Neto, J.; Palma, M. S. Characterization of Two Novel Polyfunctional Mastoparan Peptides from the Venom of the Social Wasp Polybia Paulista. Peptides 2009, 30, 1387−1395. (70) Charpentier, T. H.; Thompson, L. E.; Liriano, M. A.; Varney, K. M.; Wilder, P. T.; Pozharski, E.; Toth, E. A.; Weber, D. J. The Effects of CapZ Peptide (TRTK-12) Binding to S100B-Ca2+ as Examined by NMR and X-Ray Crystallography. J. Mol. Biol. 2010, 396, 1227−1243. (71) Chen, J. Intrinsically Disordered p53 Extreme C-Terminus Binds to S100B(β,β) through “Fly-Casting. J. Am. Chem. Soc. 2009, 131, 2088−2089. (72) Rust, R. R.; Baldisseri, D. M.; Weber, D. J. Structure of the Negative Regulatory Domain of p53 Bound to S100B(ββ). Nat. Struct. Biol. 2000, 7, 570−574. (73) Staneva, I.; Huang, Y.; Liu, Z.; Wallin, S. Binding of Two Intrinsically Disordered Peptides to a Multi-Specific Protein: A Combined Monte Carlo and Molecular Dynamics Study. PLoS Comput. Biol. 2012, 8, 1−9. (74) Case, D. A.; Babin, V.; Berryman, J. T.; Betz, R. M.; Cai, Q.; Cerutti, D. S.; T.E. Cheatham, I.; Darden, T. A.; Duke, R. E.; Gohlke, H.; et al. AMBER 14; University of California: San Francisco, 2014. (75) C. Cézard, E. Vanquelef, J. Pecher, P. Sonnet, P. Cieplak, E. D.; Dupradeau, F.-Y. RESP Charge Derivation and Force Field Topology Database Generation for Complex Bio-Molecular Systems and Analogs; 236th ACS National Meeting, Philadelphia, PA, U.S.A., August 17, 2008. 726

DOI: 10.1021/acs.jctc.5b01211 J. Chem. Theory Comput. 2016, 12, 714−727

Article

Journal of Chemical Theory and Computation (76) Patriksson, A.; van der Spoel, D. A Temperature Predictor for Parallel Tempering Simulations. Phys. Chem. Chem. Phys. 2008, 10, 2073−2077. (77) Kabsch, W.; Sander, C. Dictionary of Protein Secondary Structure: Pattern Recognition of Hydrogen-Bonded and Geometrical Features. Biopolymers 1983, 22, 2577−2637. (78) Humphrey, W.; Dalke, A.; Schulten, K. VMD: Visual Molecular Dynamics. J. Mol. Graphics 1996, 14, 33−38. (79) Toniolo, C.; Maria, G.; Vincenzo, B.; Bavoso, A.; Benedetti, E.; Blasio, B.; Di Grimaldi, P.; Lelj, F.; Pavone, V.; Pedoneb, C.; et al. Conformation of Pleionomers of A-Aminoisobutyric Acid. Macromolecules 1985, 18, 895−902. (80) Longo, E.; Moretto, A.; Formaggio, F.; Toniolo, C. The Critical Main-Chain Length for Helix Formation in Water: Determined in a Peptide Series with Alternating Aib and Ala Residues Exclusively and Detected with ECD Spectroscopy. Chirality 2011, 23, 756−760. (81) Karle, I. L. Controls Exerted by the Aib Residue: Helix Formation and Helix Reversal. Biopolymers 2001, 60, 351−365. (82) García, A. E.; Sanbonmatsu, K. Y. A-Helical Stabilization by Side Chain Shielding of Backbone Hydrogen Bonds. Proc. Natl. Acad. Sci. U. S. A. 2002, 99, 2782−2787. (83) Freddolino, P. L.; Liu, F.; Gruebele, M.; Schulten, K. TenMicrosecond Molecular Dynamics Simulation of a Fast-Folding WW Domain. Biophys. J. 2008, 94, L75−L77. (84) Pak, Y.; Kim, E.; Jang, S. Misfolded Free Energy Surface of a Peptide with Alphabetabeta Motif (1PSV) Using the Generalized Born Solvation Model. J. Chem. Phys. 2004, 121, 9184−9185. (85) Geney, R.; Layten, M.; Gomperts, R.; Hornak, V.; Simmerling, C. Investigation of Salt Bridge Stability in a Generalized Born Solvent Model. J. Chem. Theory Comput. 2006, 2, 115−127. (86) Venkatraman, J.; Shankaramma, S. C.; Balaram, P. Design of Folded Peptides. Chem. Rev. 2001, 101, 3131−3152. (87) Rubinstein, M.; Niv, M. Y. Peptidic Modulators of ProteinProtein Interactions: Progress and Challenges in Computational Design. Biopolymers 2009, 91, 505−513. (88) Pieraccini, S.; Saladino, G.; Cappelletti, G.; Cartelli, D.; Francescato, P.; Speranza, G.; Manitto, P.; Sironi, M. In Silico Design of Tubulin-Targeted Antimitotic Peptides. Nat. Chem. 2009, 1, 642− 648. (89) Pellegrino, S.; Ronda, L.; Annoni, C.; Contini, A.; Erba, E.; Gelmi, M. L.; Piano, R.; Paredi, G.; Mozzarelli, A.; Bettati, S. Molecular Insights into Dimerization Inhibition of c-Maf Transcription Factor. Biochim. Biophys. Acta, Proteins Proteomics 2014, 1844, 2108−2115. (90) Ahrens, V. M.; Bellmann-Sickert, K.; Beck-Sickinger, A. G. Peptides and Peptide Conjugates: Therapeutics on the Upward Path. Future Med. Chem. 2012, 4, 1567−1586.

727

DOI: 10.1021/acs.jctc.5b01211 J. Chem. Theory Comput. 2016, 12, 714−727