Limits...
Molecular evolution of a peptide GPCR ligand driven by artificial neural networks.

Bandholtz S, Wichard J, Kühne R, Grötzinger C - PLoS ONE (2012)

Bottom Line: The network training was based on gradient descent techniques which rely on the efficient calculation of the gradient by back-propagation.After three consecutive cycles of sequence design by the neural network, peptide synthesis and bioassay this new approach yielded a ligand with 70fold higher metabolic stability compared to the wild type peptide without loss of the subnanomolar activity in the biological assay.Combining specialized neural networks with an exploration of the combinatorial amino acid sequence space by genetic algorithms represents a novel rational strategy for peptide design and optimization.

View Article: PubMed Central - PubMed

Affiliation: Charité-Universitätsmedizin Berlin, Campus Virchow-Klinikum, Department of Hepatology and Gastroenterology and Molecular Cancer Research Center (MKFZ), Tumor Targeting Lab, Berlin, Germany.

ABSTRACT
Peptide ligands of G protein-coupled receptors constitute valuable natural lead structures for the development of highly selective drugs and high-affinity tools to probe ligand-receptor interaction. Currently, pharmacological and metabolic modification of natural peptides involves either an iterative trial-and-error process based on structure-activity relationships or screening of peptide libraries that contain many structural variants of the native molecule. Here, we present a novel neural network architecture for the improvement of metabolic stability without loss of bioactivity. In this approach the peptide sequence determines the topology of the neural network and each cell corresponds one-to-one to a single amino acid of the peptide chain. Using a training set, the learning algorithm calculated weights for each cell. The resulting network calculated the fitness function in a genetic algorithm to explore the virtual space of all possible peptides. The network training was based on gradient descent techniques which rely on the efficient calculation of the gradient by back-propagation. After three consecutive cycles of sequence design by the neural network, peptide synthesis and bioassay this new approach yielded a ligand with 70fold higher metabolic stability compared to the wild type peptide without loss of the subnanomolar activity in the biological assay. Combining specialized neural networks with an exploration of the combinatorial amino acid sequence space by genetic algorithms represents a novel rational strategy for peptide design and optimization.

Show MeSH

Related in: MedlinePlus

GNNs are trained and then optimized.This example illustrates the GNN's mode of action in computer-assisted peptide design: In the training set, the model is taught the properties of the current peptides (biological activity, stability) and the adopted cells build virtual peptides that are evaluated in the genetic algorithm-based optimization. The peptide optimization process is organized in multiple consecutive cycles. The start population of peptides is based on experts' knowledge concerning the target, e.g. natural ligands, known analogs or compounds that bind to similar targets. The trained GNN is used as a fitness function in a genetic algorithm. Newly generated sequences then have to be synthesized and analyzed in biological assays, before the next GNN training is initiated.
© Copyright Policy
Related In: Results  -  Collection


getmorefigures.php?uid=PMC3351444&req=5

pone-0036948-g002: GNNs are trained and then optimized.This example illustrates the GNN's mode of action in computer-assisted peptide design: In the training set, the model is taught the properties of the current peptides (biological activity, stability) and the adopted cells build virtual peptides that are evaluated in the genetic algorithm-based optimization. The peptide optimization process is organized in multiple consecutive cycles. The start population of peptides is based on experts' knowledge concerning the target, e.g. natural ligands, known analogs or compounds that bind to similar targets. The trained GNN is used as a fitness function in a genetic algorithm. Newly generated sequences then have to be synthesized and analyzed in biological assays, before the next GNN training is initiated.

Mentions: GNN training was initiated by feeding in the peptides' current sequences and biochemical properties such as agonistic activity and metabolic stability (Fig. 2). The training procedure was based on stochastic gradient descent with several improvements that make the training of the shared weights feasible [53]. The application of GNN-based molecular optimization was organized in a circular fashion (Fig. 2). From a start population of molecule sequences derived from established knowledge, peptides were synthesized and their properties were determined using the appropriate biochemical or cellular assay.


Molecular evolution of a peptide GPCR ligand driven by artificial neural networks.

Bandholtz S, Wichard J, Kühne R, Grötzinger C - PLoS ONE (2012)

GNNs are trained and then optimized.This example illustrates the GNN's mode of action in computer-assisted peptide design: In the training set, the model is taught the properties of the current peptides (biological activity, stability) and the adopted cells build virtual peptides that are evaluated in the genetic algorithm-based optimization. The peptide optimization process is organized in multiple consecutive cycles. The start population of peptides is based on experts' knowledge concerning the target, e.g. natural ligands, known analogs or compounds that bind to similar targets. The trained GNN is used as a fitness function in a genetic algorithm. Newly generated sequences then have to be synthesized and analyzed in biological assays, before the next GNN training is initiated.
© Copyright Policy
Related In: Results  -  Collection

Show All Figures
getmorefigures.php?uid=PMC3351444&req=5

pone-0036948-g002: GNNs are trained and then optimized.This example illustrates the GNN's mode of action in computer-assisted peptide design: In the training set, the model is taught the properties of the current peptides (biological activity, stability) and the adopted cells build virtual peptides that are evaluated in the genetic algorithm-based optimization. The peptide optimization process is organized in multiple consecutive cycles. The start population of peptides is based on experts' knowledge concerning the target, e.g. natural ligands, known analogs or compounds that bind to similar targets. The trained GNN is used as a fitness function in a genetic algorithm. Newly generated sequences then have to be synthesized and analyzed in biological assays, before the next GNN training is initiated.
Mentions: GNN training was initiated by feeding in the peptides' current sequences and biochemical properties such as agonistic activity and metabolic stability (Fig. 2). The training procedure was based on stochastic gradient descent with several improvements that make the training of the shared weights feasible [53]. The application of GNN-based molecular optimization was organized in a circular fashion (Fig. 2). From a start population of molecule sequences derived from established knowledge, peptides were synthesized and their properties were determined using the appropriate biochemical or cellular assay.

Bottom Line: The network training was based on gradient descent techniques which rely on the efficient calculation of the gradient by back-propagation.After three consecutive cycles of sequence design by the neural network, peptide synthesis and bioassay this new approach yielded a ligand with 70fold higher metabolic stability compared to the wild type peptide without loss of the subnanomolar activity in the biological assay.Combining specialized neural networks with an exploration of the combinatorial amino acid sequence space by genetic algorithms represents a novel rational strategy for peptide design and optimization.

View Article: PubMed Central - PubMed

Affiliation: Charité-Universitätsmedizin Berlin, Campus Virchow-Klinikum, Department of Hepatology and Gastroenterology and Molecular Cancer Research Center (MKFZ), Tumor Targeting Lab, Berlin, Germany.

ABSTRACT
Peptide ligands of G protein-coupled receptors constitute valuable natural lead structures for the development of highly selective drugs and high-affinity tools to probe ligand-receptor interaction. Currently, pharmacological and metabolic modification of natural peptides involves either an iterative trial-and-error process based on structure-activity relationships or screening of peptide libraries that contain many structural variants of the native molecule. Here, we present a novel neural network architecture for the improvement of metabolic stability without loss of bioactivity. In this approach the peptide sequence determines the topology of the neural network and each cell corresponds one-to-one to a single amino acid of the peptide chain. Using a training set, the learning algorithm calculated weights for each cell. The resulting network calculated the fitness function in a genetic algorithm to explore the virtual space of all possible peptides. The network training was based on gradient descent techniques which rely on the efficient calculation of the gradient by back-propagation. After three consecutive cycles of sequence design by the neural network, peptide synthesis and bioassay this new approach yielded a ligand with 70fold higher metabolic stability compared to the wild type peptide without loss of the subnanomolar activity in the biological assay. Combining specialized neural networks with an exploration of the combinatorial amino acid sequence space by genetic algorithms represents a novel rational strategy for peptide design and optimization.

Show MeSH
Related in: MedlinePlus