Deep learning is playing a vital role in every field which involves data. It has emerged as a strong and efficient framework that can be applied to a broad spectrum of complex learning problems which were difficult to solve using traditional machine learning techniques in the past. In this study we focused on classification of protein sequences with deep learning techniques. The study of amino acid sequence is vital in life sciences. We used different word embedding techniques from Natural Language processing to represent the amino acid sequence as vectors. Our main goal was to classify sequences to four group of classes, that are DNA, RNA, Protein and hybrid. After several tests we have achieved almost 99% of train and test accuracy. We have experimented on CNN, LSTM, Bidirectional LSTM, and GRU.
National Chengchi University（国立政治大学）；University of Swat（斯瓦特大学）；University of Waterloo（滑铁卢大学）；National Taipei University of Technology（国立台北科技大学）
The classification of amino acids and their sequence analysis plays a vital role in life sciences and is a challenging task. This article uses and compares state-of-the-art deep learning models like convolution neural networks (CNN), long short-term memory (LSTM), and gated recurrent units (GRU) to solve macromolecule classification problems using amino acids. These models have efficient frameworks for solving a broad spectrum of complex learning problems compared to traditional machine learning techniques. We use word embedding to represent the amino acid sequences as vectors. The CNN extract…
De novo peptide sequencing aims to recover amino acid sequences of a peptide from tandem mass spectrometry (MS) data. Existing approaches for de novo analysis enumerate MS evidence for all amino acid classes during inference. It leads to over-trimming on receptive fields of MS data and restricts MS evidence associated with following undecoded amino acids. Our approach, DPST, circumvents these limitations with two key components: (1) A confidence value aggregation encoder to sketch spectrum representations according to amino-acid-based connectivity among MS; (2) A global-local fusion decoder to…
The correlations of primary and secondary structures were analyzed using proteins with known structure from Protein Data Bank. The correlation values of amino acid type and the eight secondary structure types at distant position were calculated for distances between -25 and 25. Shapes of the diagrams indicate that amino acids polarity and capability for hydrogen bonding have influence on the secondary structure at some distances. Clear preference of most of the amino acids towards certain secondary structure type classifies amino acids into four groups: alpha-helix admirers, strand admirers, t…
A pattern Recognition of a probability distribution of amino acids is obtained for selected families of proteins. The mathematical model is derived from a theory of protein families formation which is derived from application of a Pauli's master equation method.
In this work it is shown that three pairs of the factors appear to be the key, i.e. main factors of a natural classification of protein (canonical) amino acids within the amino acid (genetic) code. First pair: the factors of the habit of an amino acid molecule (size and polarity). Second pair: the factors of the association (type of the amino acid/enzyme reactivity and degree of the hydrophobicity/hydrophilicity of an amino acid molecule). Third pair: the factors of the dissociation (degree of the acidity-basicity, over acidic group, COOH and degree of the basicity/acidity over the basic group…
Methods for alignment of protein sequences typically measure similarity by using substitution matrix with scores for all possible exchanges of one amino acid with another. Although widely used, the matrices derived from homologous sequence segments, such as Dayhoff's PAM matrices and Henikoff's BLOSUM matrices, are not specific for protein conformation identification. Using a different approach, we got many amino acid segment blocks. For each of them, the protein secondary structure is identical. Based on these blocks, we have derived new amino acid substitution matrices. The application of th…
This note represents the further progress in understanding the determination of the genetic code by Golden mean (Rakocevic, 1998). Three classes of amino acids that follow from this determination (the 7 "golden" amino acids, 7 of their complements, and 6 non-complements) are observed now together with two further possible splittings into 4 x 5 and 5 x 4 amino acids.
The twenty protein coding amino acids are found in proteomes with different relative abundances. The most abundant amino acid, leucine, is nearly an order of magnitude more prevalent than the least abundant amino acid, cysteine. Amino acid metabolic costs differ similarly, constraining their incorporation into proteins. On the other hand, sequence diversity is necessary for protein folding, function and evolution. Here we present a simple model for a cost-diversity trade-off postulating that natural proteomes minimize amino acid metabolic flux while maximizing sequence entropy. The model expla…
The observed correlations between pairs of homologous protein sequences are typically explained in terms of a Markovian dynamic of amino acid substitution. This model assumes that every location on the protein sequence has the same background distribution of amino acids, an assumption that is incompatible with the observed heterogeneity of protein amino acid profiles and with the success of profile multiple sequence alignment. We propose an alternative model of amino acid replacement during protein evolution based upon the assumption that the variation of the amino acid background distribution…
In several previous works, I presented the mirror symmetry in the set of protein amino acids, expressed through the number of atoms. Here, however, the same thing is shown but over the number of nucleons and molecules mass. Compared to the previous version of the paper, minimal changes have been made, and Display 2 as well as Figures 3 and 4 have been added.
We determine potentials of the mean force for interactions of amino acids with four common surfaces of ZnO in aqueous solutions. The method involves all-atom molecular dynamics simulations combined with the umbrella sampling technique. The profiled nature of the density of water with the strongly adsorbed first layer affects the approach of amino acids to the surface and generates either repulsion or weak binding. The largest binding energy is found for tyrosine interacting with the surface in which the Zn ions are at the top. It is equal to 7 kJ/mol which is comparable to that of the hydrogen…
Of the twenty amino acids used in proteins, ten were formed in Miller's atmospheric discharge experiments. The two other major proposed sources of prebiotic amino acid synthesis include formation in hydrothermal vents and delivery to Earth via meteorites. We combine observational and experimental data of amino acid frequencies formed by these diverse mechanisms and show that, regardless of the source, these ten early amino acids can be ranked in order of decreasing abundance in prebiotic contexts. This order can be predicted by thermodynamics. The relative abundances of the early amino acids w…
We calculate potentials of the mean force for twenty amino acids in the vicinity of the (111) surface of gold, for several dipeptides, and for some analogs of the side chains, using molecular dynamics simulations and the umbrella sampling method. We compare results obtained within three different force fields: one hydrophobic (for a contaminated surface) and two hydrophilic. All of these fields lead to good binding with very different specificities and different patterns in the density and polarization of water. The covalent bond with the sulfur atom on cysteine is modeled by the Morse potenti…
A deep neural network based architecture was constructed to predict amino acid side chain conformation with unprecedented accuracy. Amino acid side chain conformation prediction is essential for protein homology modeling and protein design. Current widely-adopted methods use physics-based energy functions to evaluate side chain conformation. Here, using a deep neural network architecture without physics-based assumptions, we have demonstrated that side chain conformation prediction accuracy can be improved by more than 25%, especially for aromatic residues compared with current standard method…
Univ. Politècnica de Catalunya（加泰罗尼亚理工大学）；UPC BarcelonaTech（UPC巴塞罗那技术学院）；Intelligent Data Science and Artificial Intelligence (IDEAI-UPC) Research Center（智能数据科学与人工智能（IDEAI-UPC）研究中心）
G-Protein Coupled Receptors (GPCRs) are a big family of eukaryotic cell transmembrane proteins, responsible for numerous biological processes. From a practical viewpoint around 34\% of the drugs approved by the US Food and Drug Administration target these receptors. They can be analyzed from their simulated molecular dynamics, including the prediction of their behavior in the presence of drugs. In this paper, the capability of Long Short-Term Memory Networks (LSTMs) are evaluated to learn and predict the molecular dynamic trajectories of a receptor. Several models were trained with the 3D posi…
Pure and homogeneous biological macromolecules (i.e. proteins, nucleic acids, protein-protein or protein-nucleic acid complexes, and functional assemblies such as ribosomes and viruses) are the key for consistent and reliable biochemical and biophysical measurements, as well as for reproducible crystallizations, best crystal diffraction properties, and exploitable electron microscopy images. Highlights: Pure and homogeneous macromolecules are the key for the best experimental results; They warrant the consistency and the reliability of biochemical and biophysical data; They give more reproduci…
The possibility of deriving the contact potentials between amino acids from their frequencies of occurence in proteins is discussed in evolutionary terms. This approach allows the use of traditional thermodynamics to describe such frequencies and, consequently, to develop a strategy to include in the calculations correlations due to the spatial proximity of the amino acids and to their overall tendency of being conserved in proteins. Making use of a lattice model to describe protein chains and defining a "true" potential, we test these strategies by selecting a database of folding model sequen…
In order to extend the results obtained with minimal lattice models to more realistic systems, we study a model where proteins are described as a chain of 20 kinds of structureless amino acids moving in a continuum space and interacting through a contact potential controlled by a 20x20 quenched random matrix. The goal of the present work is to design and characterize amino acid sequences folding to the SH3 conformation, a 60-residues recognition domain common to many regulatory proteins. We show that a number of sequences can fold, starting from a random conformation, to within a distance root…
Upon the covalent-bonding hybrid of the nitrogen atoms taken as a measure for the structural regularity in nucleobases, it can be identified that the internal relation within the 20 amino acids follows a cooperative vector-in-space addition principle based on the spherical and rotational symmetry of a quasi-28-gon (quasi-icosikaioctagon), with two evolutionary axes.
Many commonly used force fields for protein systems such as AMBER, CHARMM, GROMACS, OPLS, and ECEPP have amino-acid-independent force-field parameters of main-chain torsion-energy terms. Here, we propose a new type of amino-acid-dependent torsion-energy terms in the force fields. As an example, we applied this approach to AMBER ff03 force field and determined new amino-acid-dependent parameters for $ψ$ and $ψ'$ angles for each amino acid by using our optimization method, which is one of the knowledge-based approach. In order to test the validity of the new force-field parameters, we then perfo…
Residue-wise contact order (RWCO) is a new kind of one-dimensional protein structures which represents the extent of long-range contacts. We have recently shown that a set of three types of one-dimensional structures (secondary structure, contact number, and RWCO) contains sufficient information for reconstructing the three-dimensional structure of proteins. Currently, there exist prediction methods for secondary structure and contact number from amino acid sequence, but none exists for RWCO. Also, the properties of amino acids that affect RWCO is not clearly understood. Here, we present a lin…
We report on the successful synthesis and hyperpolarization of N unprotected α amino acid ethyl acrylate esters and extensively, on an alanine derivative hyperpolarized by PHIP (4.4$\pm$1% $^{13}$C-polarization), meeting required levels for in vivo detection. Using water as solvent increases biocompatibility and the absence of N-protection is expected to maintain biological activity.
The growing usage of nanoparticles of zinc sulfide as quantum dots and biosensors calls for a theoretical assessment of interactions of ZnS with biomolecules. We employ the molecular-dynamics-based umbrella sampling method to determine potentials of mean force for 20 single amino acids near the ZnS (110) surface in aqueous solutions. We find that five amino acids do not bind at all and the binding energy of the remaining amino acids does not exceed 4.3 kJ/mol. Such energies are comparable to those found for ZnO (and to hydrogen bonds in proteins) but the nature of the specificity is different.…
The native three dimensional structure of a single protein is determined by the physico chemical nature of its constituent amino acids. The twenty different types of amino acids, depending on their physico chemical properties, can be grouped into three major classes - hydrophobic, hydrophilic and charged. We have studied the anatomy of the weighted and unweighted networks of hydrophobic, hydrophilic and charged residues separately for a large number of proteins. Our results show that the average degree of the hydrophobic networks has significantly larger value than that of hydrophilic and char…
We derive structural and binding energy trends for twenty amino acids, their dipeptides, and their interactions with the divalent cations Ca$^{2+}$, Ba$^{2+}$, Sr$^{2+}$, Cd$^{2+}$, Pb$^{2+}$, and Hg$^{2+}$. The underlying data set consists of 45,892 first-principles predicted conformers with relative energies up to about 4 eV (about 400kJ/mol). We show that only very few distinct backbone structures of isolated amino acids and their dipeptides emerge as lowest-energy conformers. The isolated amino acids predominantly adopt structures that involve an acidic proton shared between the carboxy an…
Biosyntheses of proteins, nucleotides and fatty acids, are essential for the malignant proliferation and survival of cancer cells. Cumulating research findings show that amino acid restrictions are potential strategies for cancer interventions. Meanwhile, dietary strategies are popular among cancer patients. However, there is still lacking solid rationale to clarify what is the best strategy, why and how it is. Here, integrated analyses and comprehensive summaries for the abundances, signalling and functions of amino acids in proteomes, metabolism, immunity and food compositions, suggest that,…
In the present work, we review the fundamental methods which have been developed in the last few years for classifying into families and clans the distribution of amino acids in protein databases. This is done through functions of random variables, the Entropy Measures of probabilities of occurrence of the amino acids. An intensive study of the Pfam databases is presented with restrictions to families which could be represented by rectangular arrays of amino acids with m rows (protein domains) and n columns (amino acids). This work is also an invitation to scientific research groups worldwide …
A general strategy is described for finding which amino acid sequences have native states in a desired conformation (inverse design). The approach is used to design sequences of 48 hydrophobic and polar aminoacids on three-dimensional lattice structures. Previous studies employing a sequence-space Monte-Carlo technique resulted in the successful design of one sequence in ten attempts. The present work also entails the exploration of conformations that compete significantly with the target structure for being its ground state. The design procedure is successful in all the ten cases.
A two amino acid (hydrophobic and polar) scheme is used to perform the design on target conformations corresponding to the native states of twenty single chain proteins. Strikingly, the percentage of successful identification of the nature of the residues benchmarked against naturally occurring proteins and their homologues is around 75 % independent of the complexity of the design procedure. Typically, the lowest success rate occurs for residues such as alanine that have a high secondary structure functionality. Using a simple lattice model, we argue that one possible shortcoming of the model…
Using molecular dynamics simulations, the structure of model mini-protein was thoroughly characterized in the imidazolium-based amino acid ionic liquids and their aqueous solutions. We report that the mini-protein is more stable when AAIL is added as a cosolvent. Complete substitution of water by organic cations and anions further results in hindered conformational flexibility of the mini-protein. This observation suggests that AAILs are able to defend proteins from thermally induced denaturation. We show by means of radial distributions that the mini-protein is efficiently solvated by both so…
Cell-free protein synthesis (CFPS) systems are an attractive to complement the usual cell-based synthesis of proteins, especially for screening approaches. The literature describes a wide variety of CFPS systems, but their performance is difficult to compare since the reaction components are often used at different concentrations. Therefore, we have developed a calculation tool based on amino acid balancing to evaluate the performance of CFPS by determining the fractional yield as the ratio between theoretically achievable and achieved protein molar concentration. This tool was applied to a se…
A mechanism for creating amino acid enantiomerism that always selects the same global chirality is identified, and subsequent chemical replication and galactic mixing that would populate the galaxy with the predominant species is described. This involves: (1) the spin of the 14N in the amino acids, or in precursor molecules from which amino acids might be formed, coupling to the chirality of the molecules; 2) the neutrinos emitted from the supernova, together with magnetic field from the nascent neutron star or black hole formed from the supernova selectively destroying one orientation of the …
The present study reports on the bio-organic composition of a deep-sea venting hydrothermal system originating from arc volcanism; the origin of the particulates in hydrothermal fluids from the Suiyo Seamount in the southern Izu-Bonin (Ogasawara) Arc is discussed with regard to amino compounds. Chimney samples on deep-sea hydrothermal systems and core samples at Suiyo Seamount were determined for amino acids, and occasionally amino sugars. Two types of chimney samples were obtained from active hydrothermal systems by submersible vehicles: one was natural chimney (NC) on a hydrothermal natural …
The information regarding the structure of a single protein is encoded in the network of interacting amino acids. Considering each protein as a weighted and unweighted network of amino acids we have analyzed a total of forty nine protein structures that covers the three branches of life on earth. Our results show that the probability degree distribution of network connectivity follows Poisson's distribution; whereas the probability strength distribution does not follow any known distribution. However, the average strength of amino acid node depends on its degree (k). For some of the proteins, …
Topological properties of native folds are obtained from statistical analysis of 160 low homology proteins covering the four structural classes. This is done analysing one, two and three-vertex joint distribution of quantities related to the corresponding network of amino acid residues. Emphasis on the amino acid residue hydrophobicity leads to the definition of their center of mass as vertices in this contact network model with interactions represented by edges. The network analysis helps us to interpret experimental results such as hydrophobic scales and fraction of buried accessible surface…
All known terrestrial proteins are coded as continuous strings of ~20 amino acids. The patterns formed by the repetitions of elements in groups of finite sequences describes the natural architectures of protein families. We present a method to search for patterns and groupings of patterns in protein sequences using a mathematically precise definition for 'repetition', an efficient algorithmic implementation and a robust scoring system with no adjustable parameters. We show that the sequence patterns can be well-separated into disjoint classes according to their recurrence in nested structures.…
We study the coupled dynamics of primary and secondary structure formation (i.e. slow genetic sequence selection and fast folding) in the context of a solvable microscopic model that includes both short-range steric forces and and long-range polarity-driven forces. Our solution is based on the diagonalization of replicated transfer matrices, and leads in the thermodynamic limit to explicit predictions regarding phase transitions and phase diagrams at genetic equilibrium. The predicted phenomenology allows for natural physical interpretations, and finds satisfactory support in numerical simulat…
Background:Prediction of protein three-dimensional structures from amino acid sequences is a long-standing goal in computational/molecular biology. The successful discrimination of protein folds would help to improve the accuracy of protein 3D structure prediction. Results: In this work, we propose a method based on linear discriminant analysis (LDA) for recognizing proteins belonging to 30 different folds using the occurrence of amino acid residues in a set of 1612 proteins. The present method could discriminate the globular proteins from 30 major folding types with the sensitivity of 37%, wh…
Atlantic sea lamprey contains two corticoid receptors (CRs), CR1 and CR2, that are identical except for a four amino acid insert (Thr-Arg-Gln-Gly) in the CR1 DNA-binding domain (DBD). Steroids are stronger transcriptional activators of CR2 than of CR1 suggesting that the insert reduces the transcriptional response of lamprey CR1 to steroids. The DBD in elephant shark mineralocorticoid receptor (MR) and glucocorticoid receptor (GR), which are descended from a CR, lack these four amino acids, suggesting that a CR2 is their common ancestor. To determine if, similar to lamprey CR1, the presence of…
What are proteins made from, as the working parts of the living cells protein machines? To answer this question, we need a technology to disassemble proteins onto elementary func-tional details and to prepare lumped description of such details. This lumped description might have a multiple material realization (in amino acids). Our hypothesis is that informational approach to this problem is possible. We propose a way of hierarchical classification that makes the primary structure of protein maximally non-random. The first steps of the suggested research program are realized: the method and th…
We present a structural data set of the 20 proteinogenic amino acids and their amino-methylated and acetylated (capped) dipeptides. Different protonation states of the backbone (uncharged and zwitterionic) were considered for the amino acids as well as varied side chain protonation states. Furthermore, we studied amino acids and dipeptides in complex with divalent cations (Ca2+, Ba2+, Sr2+, Cd2+, Pb2+, and Hg2+). The database covers the conformational hierarchies of 280 systems in a wide relative energy range of up to 4 eV (390 kJ/mol), summing up to an overall of 45,892 stationary points on t…
This article is interested in the origin of the genetic code, it puts forward a scenario of a simultaneous selection of the bases and amino acids and setting up of a correlation between them. Each amino acid is associated with a pair of its own kind, called the binding pair and each binding pair is associated with the codon(s) corresponding to the same amino acid. An explanation is also proposed about the origin of the start and stop codons.
We derive an analytic expression for site-specific stationary distributions of amino acids from the Structurally Constrained Neutral (SCN) model of protein evolution with conservation of folding stability. The stationary distributions that we obtain have a Boltzmann-like shape, and their effective temperature parameter, measuring the limit of divergent evolutionary changes at a given site, can be predicted from a site-specific topological property, the principal eigenvector of the contact matrix of the native conformation of the protein. These analytic results, obtained without free parameters…
Learning how proteins fold will hardly have any impact in the way conventional -- active site centered -- drugs are designed. On the other hand, this knowledge is proving instrumental in defining a new paradigm for the identification of drugs against any target protein: folding inhibition. Targeting folding renders drugs less prone to elicit spontaneous genetic mutations which in many cases, notably in connection with viruses like the Human Immunodeficiency Virus (HIV), can block therapeutic action. From the progress which has taken place during the last years in the understanding of the becom…
The amino acid sequences of proteins provide rich information for inferring distant phylogenetic relationships and for predicting protein functions. Estimating the rate matrix of residue substitutions from amino acid sequences is also important because the rate matrix can be used to develop scoring matrices for sequence alignment. Here we use a continuous time Markov process to model the substitution rates of residues and develop a Bayesian Markov chain Monte Carlo method for rate estimation. We validate our method using simulated artificial protein sequences. Because different local regions s…
Extensive molecular dynamics simulations reveal that the interactions between proteins and poly(ethylene glycol)(PEG) can be described in terms of the surface composition of the proteins. PEG molecules accumulate around non-polar residues while avoiding polar ones. A solvent-accessible-surface-area model of protein adsorption on PEGylated nanoparticles accurately fits a large set of data on the composition of the protein corona recently obtained by label-free proteomic mass spectrometry.
Heat-bath cooling is a component of practicable algorithmic cooling of spins, an approach which might be useful for in vivo 13C spectroscopy, in particular for prolonged metabolic processes where substrates that are hyperpolarized ex-vivo are not effective. We applied heat-bath cooling to 1,2-13C2-amino acids, using the alpha protons to shift entropy from selected carbons to the environment. For glutamate and glycine, both carbons were cooled by about 2.5-fold, and in other experiments the polarization of C1 nearly doubled while all other spins had equilibrium polarization, indicating reductio…
In this work it is shown that 20 canonical amino acids (AAs) within genetic code appear to be a whole system with strict AAs positions; more exactly, with AAs ordinal number in three variants; first variant 00-19, second 00-21 and third 00-20. The ordinal number follows from the positions of belonging codons, i.e. their digrams (or doublets). The reading itself is a reading in quaternary numbering system if four bases possess the values within a specific logical square: A = 0, C = 1, G = 2, U = 3. By this, all splittings, distinctions and classifications of AAs appear to be in accordance to at…
The in silico design of peptides and proteins as binders is useful for diagnosis and therapeutics due to their low adverse effects and major specificity. To select the most promising candidates, a key matter is to understand their interactions with protein targets. In this work, we present PARCE, an open source Protocol for Amino acid Refinement through Computational Evolution that implements an advanced and promising method for the design of peptides and proteins. The protocol performs a random mutation in the binder sequence, then samples the bound conformations using molecular dynamics simu…
We introduce and numerically solve a mathematical model of the experimentally established mechanisms responsible for the symmetry breaking transition observed in the chiral crystallization experiments reported by I. Weissbuch, L. Addadi, L. Leiserowitz and M. Lahav, J. Am. Chem. Soc. 110 (1988), 561-567. The mathematical model is based on five basic processes: (1) The formation of achiral glycine clusters in solution, (2) The nucleation of oriented glycine crystals at the air/water interface in the presence of hydrophobic amino acids, (3) A kinetic orienting effect which inhibits crystal growt…
