INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS AND DISCUSSION CONCLUSIONS REFERENCES

It is well known that the biological activity depends on the spatial conformation acquired by the macromolecules in the physiological medium. The action of hormones and drugs is also dependent on the molecular three-dimensional structure of the target molecules. In recent years, the atomic description of biological molecules in computational simulations have promoted significant advances in the comprehension of the biological process as well as proposed new insights in the design of molecules to satisfy specific properties.

Biological macromolecules have a large number of degrees of freedom leading to several local minima in the molecular energy hyper-surface. Concerning protein functionality, it is presumed that these molecules express their biological activity when they are close to the global minimum of energy (Yon, 1997). A generalized concern is how to predict the lowest conformational energy through simulation procedures. In that sense, we have developed a computational code to perform stochastic molecular optimization (Moret et al., 1998b) hoping to find the lowest conformational energy.

The available literature on peptide conformational analysis is enormous, since Corey and Pauling (1953) determined the ideal values for all backbone bond lengths and bond angles. The allowed values for the pairs of dihedral angles about the C atoms, which are limited by steric constraints, were determined by Ramachandran et al. (1963) and are summarized in the so-called Ramachandran map.

Important advances are found in the theoretical field of chain topologies prediction (Rooman et al., 1991) and also regarding the folding patterns determined by experimental studies (Richards, 1991; Dobson, 1992; Elöve et al., 1992; Haynie and Freire, 1993; Kiefhaber et al., 1992; Radford et al., 1992; Khorazanizadeh et al., 1993; Evans and Radford, 1994; Clarke et al., 1999) as well as by theoretical ones (Dill, 1990, 1993; Shakhnovich, 1994; Sali et al., 1994; Wolynes et al., 1995; Bryngelson et al., 1995; Pande et al., 1998; Hiltpold et al., 2000).

Helices are the most prevalent secondary structural motif observed in proteins with known structure. Several recent theoretical studies of the helix-coil transitions have shown the relation between the unfolded state, in random coil conformations, and the helical conformational states, by using molecular dynamics (Daggett et al., 1991; Tobias and Brooks, 1991; Wang et al., 1995; Brooks, 1996, Doruker and Bahar, 1997; Bertsch et al., 1998, Hiltpold et al., 2000), Monte Carlo (Sung, 1994, 1995; Wu and Wang, 1998), or other theoretical approaches (Jun and Weaver, 2000; Park and Goddard, 2000). Both standard molecular dynamics and Monte Carlo simulation methods require extended computational time to obtain helical secondary structures.

To search peptide conformational space, we proposed a very fast stochastic procedure consisting of a simulated annealing methodology coupled to a biased search of the energy hyper-surface according to the allowed regions in the Ramachandran map. We applied this procedure to investigate the helical propensities of polyalanines as a function of the polypeptide size, in a continuum medium approach with a dielectric constant _r = 2. To illustrate the method applicability, a more complex system is presented briefly, where we search for the secondary structures in a small protein.

Simulated annealing methods have been applied successfully for the description of a variety of global optimization problems. The power of the simulated annealing methods is due to their suitability for large-scale optimization problems, especially for those in which a desired global minimum is hidden among many local minima. The first nontrivial solution in this sense was proposed to solve combinatorial problems (Kirkpatrick et al., 1983). Their algorithm strictly follows the quasi-equilibrium Boltzmann-Gibbs statistics using a Gaussian visiting distribution. Based on generalized thermostatistics (Tsallis, 1988, 1995; Curado and Tsallis, 1991), Tsallis and Stariolo (1996) proposed the generalized simulated annealing (GSA), or Tsallis machine, approach. Tsallis machine was applied in several problems, such as molecular optimization using classical methods (Moret et al., 1998b) or semi-empirical methods (Mundim and Tsallis, 1996), geophysical problems (Mundim et al., 1998), traveling salesman problem (Penna, 1995a), numerical data fitting (Penna, 1995b), and genetic algorithm (Moret et al., 1998a). GSA has been proven to be the most effective simulated annealing method when considered for optimizing combinatorial problems (Penna, 1995a) and seems to be the most effective one in simulated annealing when considered for optimizing problems in real space (Tsallis and Stariolo, 1996). Our strategy to obtain molecular conformational energies is to couple a classical force field with the GSA procedure and use a Ramachandran map as a tendency for the probable dihedral angles.

We considered as allowed regions of the Ramachandran map, those corresponding to the  and  angles values proposed in the PRELUDE software package (Rooman et al., 1991). These values were computed from comparative statistics of the backbone secondary structure for several amino acid sequences. Table 1 shows the seven possible conformations proposed by this method to predict main-chain topology. These specific angles describe the average conformation of a wide range of proteins with known backbone topology and were also employed by Dandekar and Argos (1994) to perform a polypeptide conformational search using a genetic algorithm. In our strategy we used these seven values (Table 1) to define the center of each allowed region for the dihedral coordinates. Dandekar and Argos (1994) used an empirical potential based on the hydrophobic contribution of each amino acid. Our procedure uses a classical force field that describes all atoms in the system whereas conformations are obtained by a stochastic method, i.e., GSA. The stochastic procedure is used to scan the peptide main chain oriented by Ramachandran map and also to analyze side-chain conformations.

In developing the computational code, for the GSA biased by the Ramachandran map procedure (GSARM), four aspects were considered to create an efficient calculation system. First, the energy function E() is defined in an N-dimension continuous space, where   R^N; second, transpose easily the conformational barrier to avoid trapping the system in local minima; third, the search is biased considering seven conformational regions that are defined in the Ramachandran map as proposed by Rooman et al. (1991); and, fourth, the method gives a rapid analysis of the energy hyper-surface, increasing information about Ramachandran maps in addition.

Here we present the GSARM procedure used for recovering the global minimum. We then discuss the results obtained for some molecular structures of simple polypeptides (polyalanines), and the conclusions of this research are presented.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS AND DISCUSSION CONCLUSIONS REFERENCES

The THOR program (Arêas et al., 1995; Moret et al., 1998b; Pascutti et al., 1999a,b) was developed to be a comprehensive and a flexible tool to investigate macromolecular structures of biological interest such as proteins and membranes. The computational code is based on a classical force field and considers the GROMOS parameters (van Gusteren and Berendsen, 1987) as well as the corrections from the GROMOS96 version; however, other force fields can be easily implemented. Both molecular dynamics and optimization methods are available in this program. The choice of either of these methods depends on the user needs. For example, if dynamic properties are needed, the software can perform molecular dynamics simulations. To analyze systematically the conformational energy hyper-surface or to map the global and local minima we use stochastic methods.

In the THOR program, the conformational energy of the molecule is made up of a sum of bonded and nonbonded terms (Arêas et al., 1995; Pascutti et al., 1999a,b). In this approach, only hydrogen atoms covalently bonded to oxygen or to nitrogen are considered explicitly, whereas CH₁, CH₂, and CH₃ groups are assumed to be an atomic unit. In our stochastic procedure we use a simplified version of the conformational energy, which maintains fixed bond lengths and bond angles within their ideal values. We focus our search on the dihedral angle space (Moret et al., 1998b). Therefore, we analyze the changes of the following energy function:

(1)

where E is the dihedral angle potential, E_VdW is the van der Waals potential, and E_el is the Coulomb potential term (see definitions and used parameters in Moret et al., 1998b).

The search of local and global minima and the mapping of the energy hyper-surface involves the comparison of the energies of two consecutive random conformations. The molecular geometry at steps t + 1 and t, are given by _t+1, and _t, respectively, where _t is an N-dimensional vector that contains all dihedral coordinates (N) to be optimized. Then, for two consecutive steps, they are related by

(2)

where _t is a random perturbation of the dihedral angles.

This random perturbation _t allows visiting the potential hyper-surface. Each point in Table 1 defines a central point of seven allowed square regions. To define the extension of these regions we have initially taken large  and  angular variations centered on each point. By testing the methodology on polyalanines, we have verified that most of the optimized structures present variations of the  and  values that did not exceed 3.6° according to the expected central points of Table 1. Therefore, we have considered the conformational search in the GSARM procedure restricted to these seven square areas of side equal to 7.2. To compare the number of cycles necessary for convergence, we have also performed conformational search on the entire  ×  space (full GSA procedure).

To generate the random vector _t+1 we use the visiting distribution function g(_t,i) for each component _t,i of the perturbation vector _t, defined as follows:

(3)

where q_v is the visiting index, T_q is the visiting temperature, and  is the gamma function. To obtain the _t vector we took a random value for the random vector a = {a_i} (0 < a_i < 1) and performed a numerical integration of the visiting distribution probability:

(4)

The integral of g_qv(_t) has an analytical solution only for q_v = 1 or q_v = 2. In the GSA package, the integral is calculated by using a series expansion and by taking the inverse function of a polynomial series, whose expansion has a cutoff at the 17th order, as proposed in Moret et al. (1998b).

The following step is to compare the energy of the new conformation related to the old one; if this energy is lower than the previous one, the new conformation is accepted. To test the acceptance of a new conformation with higher energy we use the generalized Tsallis statistic P_qA, given by Tsallis and Stariolo (1996):

(5)

where q_A is acceptance index and T_qA is the acceptance temperature. At the limit, where q_A  1 we recover the Boltzmann-Gibbs statistics used in the conventional Metropolis criterion (Metropolis et al., 1953).

In summary, the algorithm for searching the peptide secondary structure, using GSA or GSARM procedure, is as follows. 1) Fix the parameters (q_A, q_v). Start, at t = 1, with arbitrary internal coordinates (arbitrary dihedral angles) and a high enough value T₀ for the initial visiting and acceptance temperatures: T_qv(1) and T_qA(1). 2) Randomly choose a region of the Ramachandran map for each residue using the visiting distribution probability g_qv(_t) (Eq. 5). 3) Randomly generate the vector _t+1, given by the visiting distribution probability g_qv(_t) (Eq. 5). 4) Calculate the conformational energy E(_t+1) by using the THOR program and the acceptance criterion as follows: if E(_t+1) < E(_t), replace _t by _t+1; if E(_t+1)  E(_t), run a random number r  [0, 1]; if r > P_qA (acceptance probability) retain _t; otherwise, replace _t by _t+1. 5) Cool the system by decreasing the temperatures T_qv(t) and T_qA(t), assuming, for the sake of simplicity, that T_qv(t) = T_qA(t):

(6)

where t is the discrete computational step. 6) Go back to step 2 until the energy E_min() is reached, i.e., when a lower energy value is no longer obtained in several steps. A more general test for the convergence is whether the same final conformation is obtained starting from different initial conditions.

For a sufficiently large number of steps, this procedure assures that the system can escape from any local minimum and explore the entire allowed energy hyper-surface. In the following section the applications for the GSA and GSARM approaches is shown.

To study the helix formation and to compare the two methodologies (GSA and GSARM) we performed a search of minima in the energy hyper-surface of the polyalanines with 5-20 alanine residues. To obtain the dihedral minima structures we have used both GSA and GSARM approaches taking as initial conformations the completely extended ((, ) = (180^o, 180^o)) conformation for all polyalanines. In the calculations we have fixed all bond lengths and bond angles within their ideal values, as proposed by Corey and Pauling (1953). We have performed a set of simulations with different initial parameters, i.e., T₀ = [1, 2, 5, 10, 50, 100] and q_A and q_v values in the interval [1.1, 2.5] with 0.1 as a step. We observed that global minimum is reached in more than 50% of the simulations using T₀ = 100, q_A = 1.1, and q_v values larger or equal to 2. For each peptide system we performed 180 simulations (90 with GSA and 90 with GSARM). Therefore we performed 2880 simulations (1440 to GSA and 1440 to GSARM). Using this procedure we obtained sets of configurations in local and global minima.

	RESULTS AND DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS AND DISCUSSION CONCLUSIONS REFERENCES

The -helix is the classic element of protein structure. Pauling et al. (1951) were the first ones to describe the -helix. They predicted it as a stable and favorable structure in proteins. All the hydrogen bonds of the -helix backbone are aligned along the helical axis with the same orientation. Because a peptide bond has a dipole moment arising from the different polarity of the NH and CO groups, these dipole moments are also aligned along the helical axis. The overall effect is a significant macro-dipole that has the positive pole at the amino end and the negative pole at the carboxyl end of the -helix. The overall energy that stabilizes the -helix came from the attractive contributions due to hydrogen bonds and/or by charge-helix dipole interactions (Shoemaker et al., 1985) as well as from the van der Waals contributions. The -helix formation is a cooperative process where the electrostatic energy has an important role in stabilizing this type of structure, as was shown by Park and Goddard (2000) through ab initio quantum mechanics calculations. These authors found that extending the length of an -helix by adding additional residues increasingly favors the -helix formation. A critical number of amino acids, however, are necessary to stabilize this -helix structure, and an upper limit may also be imposed by the entropy effect. Isolated -helix structures, in fact, would have to be longer than 13 residues to be stabilized by the attractive interactions (Shoemaker et al., 1987; Rogers, 1989; Voet and Voet, 1994).

Different amino acids have been found to present weak though definite preference in favor or against being in -helix structure, and the intrinsic helical propensity of some amino acids has been demonstrated to be position dependent (Petukhov et al., 1999). In this sense, Ala, Glu, Leu, and Met are considered to be good -helix promoters whereas Pro, Gly, Tyr, and Ser are considered to be poor ones (Branden and Tooze, 1991). Such preferences were the main considerations in all early attempts to predict secondary structures from amino acid sequences, but they were not strong enough to obtain accurate predictions.

Short polypeptides and individual protein fragments generally do not form helices in water. The first example resulting in a significant -helix formation in water, near 0°C, was obtained with the C- and S-peptide fragments of ribonuclease A (Kim and Baldwin, 1984). Stable -helix formation was observed in 16-residue alanine-based peptides and was attributed to the high helix-forming potential of alanines (Marqusee et al., 1989). Other experimental measurements on alanine-based peptides have shown that helix nucleation takes place on the millisecond time scale (Clarke et al., 1999).

Molecular dynamics simulations, considering aqueous or implicit solvent, have also tested the stability of polyalanines. Doruker and Bahar (1997) studied the -helix stability in homopeptides of 13 amino acid residues, and they proposed a rank for the amino acids involved where Ala < Val < Ser < Gly. They observed that polyalanine unfolds within few hundreds of picoseconds at 350 K. Starting from all-coil conformations, Hiltpold et al. (2000) observed the helical stabilization of alanine-based polypeptides of ~30 residues, in an implicit solvent model, within the first 30 ns, at 360 K, with an average folding time of 10 ns. By means of torsion-coordinates molecular dynamics, a method that eliminates bond and angle degrees of freedom, Bertsch et al. (1998) observed the -helix formation in a 20-residue alanine peptide in trajectories of 0.5 ns, with a half-life of 210 ps for the helix formation. A shorter -helix folding time of 100 ps was found by Wu and Wang (1998) for a 16-residue polyalanine, with self-guided molecular dynamics, based on the motion guided by an introduced external force.

Stochastic methods have also been applied in polyalanine helix studies (Sung, 1994; Hoffmann and Knapp, 1996). Sung (1994), with the Metropolis-Monte Carlo method and the solvent-referenced interaction, observed the -helix formation in a 16-residue alanine peptide, with different initial conditions, after 15 × 10⁶ steps. In this work we applied the stochastic method GSA, and demonstrated that polyalanines are stable in the -helix structure when the peptide has 13 or more amino acid residues, in a low dielectric constant medium. Furthermore, we show that this stable conformation can be reached in a few thousand steps.

In Tables 2 and 3 are shown the 10 lowest-energy conformations obtained for most of the studied peptides, using both GSARM and GSA procedures. It can be observed that, unless a critical number of residues is attained (~13 residues), the lowest-energy state obtained corresponds to a nearly random conformation, and only few residues are in the -helix region whereas the majority of them is out of this region. Furthermore, for peptides with less than 13 residues, the energy gap between the lowest-energy state and any other lower-energy states is of the order of the available thermal energy at room temperature. On the contrary, for peptides with more than 13 residues, the lowest-energy state corresponds to a state where most of the residues are in the -helix conformation, and the energy gap between the lowest state and the following lower one tends to be very large.

We summarized our results for polyalanines in Fig. 1, showing the minimum energy conformations of some polyalanines with 5 residues up to 20 residues using these methodologies. In this figure, we note that no conformational preference is observed for peptides with less than 13 residues (Fig. 1, A-F). Peptides with 13 or more residues tended to be stabilized in an -helix structure (Fig. 1, G-J). We observe in Table 2 that the peptide with 13 alanine residues presented an -helix structure that corresponds to the lowest-energy minimum. The peptides with 6, 7, 8, 9, and 11 alanine residues presented an -helix structure corresponding only to its secondary local minimum. On the other hand, in Table 3, all peptides with more than 13 residues presented an -helix conformation corresponding to the lowest-energy minimum.

View larger version (43K): [in this window] [in a new window]   FIGURE 1   The structures correspond to the lowest minimum of the polyalanines with 5 (A), 6 (B), 7 (C), 8 (D), 9 (E), 11 (F), 13 (G), 15 (H), 18 (I), and 20 (J) residues. It is clearly shown that polyalanines with 13 or more residues have -helix conformations.

In Fig. 2 is shown the Ramachandran map for all minima energy configurations obtained with both GSARM and GSA methods for the pentalanine and for the icoalanine. It was possible to discriminate the most populated regions on the Ramachandran map. We observed that pentalanine (Fig. 2 A) populates more the -region than the -region. On the contrary, we observed that icoalanine (Fig. 2 B) populates more the -region than the -region. In Fig. 2, C-H, more details of the icoalanine simulations are shown. We noted that N-terminus residues have not been stabilized even around an -region, in accordance with the literature (Munoz and Serrano, 1995). On the other hand, the presented residues showed an -helix structure preference.

View larger version (24K): [in this window] [in a new window]   FIGURE 2   (A and B) The  and  coordinates of each residue of the minimum energy structures found by using both GSA and GSARM procedures, for pentalanine (A) and icoalanine (B). (C-H) Ramachandran plot of some residues of the icoalanine are shown in detail: N-terminus (C), ALA3 (D), ALA8 (E), ALA12 (F), ALA17 (G), and C-terminus (H).

In Fig. 3 A is shown the energy gap between the global (lowest) minimum and the first local minimum for all polyalanines. In this figure we observe that for peptides with a number of residues less than 12 residues the energy gap tends to increase with the enhancement of the peptide size. On the other hand, peptides with 13-16 residues do not present a large energy gap. Peptides with 17 or more residues presented a large energy gap, and therefore, these peptides are very stable in an -helix conformation. In Fig. 3 B is shown the energy gap between the lowest-energy state and the minimum with the largest number of the residues in the -region. In this figure we also observe that peptides with 13 or more residues tend to have an -helix conformation.

View larger version (12K): [in this window] [in a new window]   FIGURE 3   (A) Energy gap between the first local minimum and the global minimum for all polyalanines; (B) Energy gap between the structures with a largest number of residues in the -region and the lowest minimum.

Although we have used a simple version for the atomic representation of the peptide in a continuum electrostatic medium of dielectric constant _r = 2, considering the united atom model for aliphatic carbons and the GROMOS force field, the general features, i.e., polyalanine -helix preference above a critical residue number of ~13 amino acids, the energy gap from the -helix to other conformations, and the gap enhancement showing the crescent stabilization as the number of residues increases, are in good accordance with experiments (Shoemaker et al., 1987; Marqusee et al., 1989; Clarke et al., 1999) and previous theoretical predictions (Rogers, 1989; Sung, 1994; Hoffmann and Knapp, 1996; Doruker and Bahar, 1997; Bertsch et al., 1998; Wu and Wang, 1998; Park and Goddard, 2000; Hiltpold et al., 2000).

To compare both approaches, GSARM and GSA, we have analyzed the number of cycles necessary to have convergence as a function of the peptide size (Fig. 4). It is shown in Fig. 4 that the GSARM approach that took less than 1000 steps to converge is a fast method to determine peptide conformations. All energy minima obtained by both GSA and GSARM procedures for the studied polyalanines were recorded, and both procedures lead to the same results, although differences were observed in the number of interactions and consequently the computational time expended in each simulation.

View larger version (12K): [in this window] [in a new window]   FIGURE 4   The lowest number of cycles necessary to obtain a conformation of minimum energy as function of polypeptide size.

The time step in molecular dynamics simulations is of the order of 0.0005-0.0020 ps. When considering the latter, 50,000 steps are necessary for a 100-ps trajectory and millions of steps to simulate events at a nanosecond time scale. So far, it was reported in the literature that simulations of the -helix folding for a polyalanine had taken 0.1-30 ns (Wu and Wang, 1998; Bertsch et al., 1998; Hiltpold et al., 2000). Usual stochastic methods, such as Metropolis-Monte Carlo, can take millions of steps to simulate the -helix polyalanine folding (Sung, 1994). Applying the GSARM procedure, peptides as long as 20 monomers initially in an extended conformation can adopt an -helical structure in less than 1000 steps. The fact that the global energy minimum is attained in so few steps is probably due to the fractal structure of peptide and protein energy hyper-surfaces, as we have recently suggested (Moret et al., 2001).

To test our procedure on protein fragments we studied the secondary structures of the insect defensin A (Protein Data Bank code 1ICA, Cornet et al., 1995). It is a basic 4-kDa protein that in vivo is excreted in the hemolymph of the flesh fly Phormia terramovae larvae in response to bacterial challenge or tissue injury (Lambert et al., 1989) and is principally active against Gram-positive bacteria. Insect defensin A presents one -helix and two -strands stabilized by three disulfide bridges. The energy hyper-surface of the -helix and the two -sheet structures were analyzed, by means of the GSA procedure, randomly searching all the main chain  and  angles and all of the side-chain dihedral angles of these structures starting from the native conformation. The remaining residues, including the main-chain and lateral groups, were kept at the native positions.

The results for the -helix domain (HIS13-ARG23) showed that the lowest energy for the unfolded state at a random coil conformation is ~10 kcal/mol higher than the conformational energy of the -helix global minimum. However, when we performed the search in the region of -sheet (ARG26-ARG39), we found that the lowest-energy conformation corresponds to a random coil, which is only 0.8 kcal/mol less than the corresponding -sheet conformational energy.

	CONCLUSIONS

TOP ABSTRACT INTRODUCTION METHODS RESULTS AND DISCUSSION CONCLUSIONS REFERENCES

The GSA method was used to explore the energy hyper-surface of peptides. This method, based on generalized thermostatistics, is a fast stochastic approach used to determine the allowed global and local energy minima. We have applied this strategy to predict secondary structures of polyalanines. An -helix conformation was found as corresponding to the lowest conformational energy for peptides of 13 or more residues, as suggested in the literature. When the search of the peptide backbone conformations is performed with the GSARM approach, where certain values of dihedral angles  and  in the Ramachandran map are more frequently visited, the simulation generally expends less than 1000 steps to find the global minimum energy for a 5-20-residue polyalanine. Compared with other theoretical approaches based on molecular dynamics or Monte Carlo methodologies, GSARM can be considered as the fastest strategy.

The electrostatic interactions between adjacent dipoles in the peptide backbone are repulsive, so a critical number of residues is needed to be counterbalanced by the attractive interactions, such as H-bonds. Then, short peptides of less than seven residues are flexible, and they show no conformational preference. When the number of residues of the chain increases (8-13 residues) the structures tend to collapse into H-bonded turns, as shown in Fig. 1. Above a critical number of 13 amino acid residues the enhancement of the hydrogen bond number stabilizes the polyalanine in an -helix structure. In fact, for long-chain peptides, most of the possible H-bonds of the backbone tend to be formed. Then, an energy gap arises from the lowest-energy conformation to the next low-energy conformation. This energy gap enlarges with the number of residues.

The helical conformations presented in Fig. 1 were also obtained when we performed a restricted search to one quadrant of the Ramachandran map. In this case, the backbone angles  and  were restrained to the third quadrant, i.e., 180^o <  < 0^o and 180^o <  < 0^o, as proposed by Wang et al. (1995) and Tobias and Brooks (1991) to analyze helical propensities.

Finally, we performed a search considering a large value for the dielectric constant (_r = 80) to obtain the lowest-energy minimum for two systems: a 13-alanine peptide and a 14-alanine one. According our simulations these peptides tend to adopt values for their  and  dihedral angles corresponding to the -helix region in the Ramachandran map (i.e., in the third quadrant), but the H-bond formation is poorly maintained. This example as well as the simulations of the defensin show the importance of having a good description of the solvent effect, mainly when we are interested in searching for the protein conformations in its appropriated environment. This environment changes from the low dielectric constant, typical of the protein interior, to the higher one, for residues in contact with the solvent.

According to the Ramachandran map, proteins have forbidden values for the - pairs of dihedral angles. Performing the search for the peptide backbone conformations, exclusively in the allowed regions, the number of steps to achieved GSA convergence decreases drastically. Both approaches, the GSA and the GSARM, have been shown to be applicable to helix folding studies; however, the method proposed here is sufficiently general to be extended to analyze any conformational state of polypeptides and proteins in general.