Method For Prediction Of Solubility Research Articles

Research on protein solubility prediction based on ensemble learning and feature fusion

Protein solubility is one of the momentous properties of a protein that can effectively participate in and inhibit the physiological and biochemical processes of cancer cells in the human body. Therefore, understanding the solubility of proteins may be significant to find the mechanism of diseases caused by the solubility of proteins. In this paper, to improve the protein solubility prediction performance and address the inadequacy of existing protein solubility prediction methods that more feature information about protein sequences is difficult to be obtained. A protein solubility prediction model named EL-FFsol is proposed, which is based on the CatBoost ensemble learning framework and multiple feature fusion of protein sequences. First of all, protein sequence features were introduced to build fusion representation, including the Physicochemical Properties, One-hot Feature Encoding, Amino Acid Composition and Statistical Features. Additionally, the CatBoost was employed to construct an ensemble learning model to predict protein solubility. Finally, EL-FFsol was tested on the benchmark dataset to predict the solubility of proteins. In terms of accuracy, matthews correlation coefficient, sensitivity, specificity, area under ROC curve and area under P-R curve, EL-FFsol achieved 0.7679, 0.5480, 0.6630, 0.8729, 0.8540 and 0.8440 performances. Compared with the DeepSOL and DDcCNN, the matthews correlation coefficient was increased by 1.68% and 0.79%, the area under ROC curve was increased by 1.60% and 2.20% and the area under P-R curve was increased by 1.70% and 2.40%, respectively.

Dielectric relaxation and molecular interactions study of saccharides in aqueous solutions

Time domain reflectometry (TDR) technique has been extensively used to study dielectric relaxation and solution properties of carbohydrates. Using TDR techninque, complex permittivity spectra of monosaccharides (d-fructose and d-xylose) and disaccharides (d-maltose monohydrates) were obtained in the frequency range of 10 MHz–50 GHz at various concentrations and temperatures. The static dielectric constant (ε0), dielectric constant at high frequency (ε∞), relaxation time (τ) and relaxation time distribution parameter (β) extracted from the complex permittivity spectra using least squares fit method. The values of static dielectric constant were also verified by LCR meter by dielectric measurement in the frequency range of 20Hz to 2 MHz at 25 °C. The relaxation behavior of aqueous solutions of monosaccharides and disaccharides has been illustrated by using Cole-Davidson model. Activation enthalpy (ΔH), entropy (ΔS) and Kirkwood correlation factor have been determined to study extent of hydrogen bonding. This data might be useful in pharmaceutical, food processing industry and in solubility prediction method in aqueous solution.

Solubility Prediction from Molecular Properties and Analytical Data Using an In-phase Deep Neural Network (Ip-DNN).

Materials informatics is an emerging field that allows us to predict the properties of materials and has been applied in various research and development fields, such as materials science. In particular, solubility factors such as the Hansen and Hildebrand solubility parameters (HSPs and SP, respectively) and Log P are important values for understanding the physical properties of various substances. In this study, we succeeded at establishing a solubility prediction tool using a unique machine learning method called the in-phase deep neural network (ip-DNN), which starts exclusively from the analytical input data (e.g., NMR information, refractive index, and density) to predict solubility by predicting intermediate elements, such as molecular components and molecular descriptors, in the multiple-step method. For improving the level of accuracy of the prediction, intermediate regression models were employed when performing in-phase machine learning. In addition, we developed a website dedicated to the established solubility prediction method, which is freely available at “http://dmar.riken.jp/matsolca/”.

Neural network modeling based double-population chaotic accelerated particle swarm optimization and diffusion theory for solubility prediction

Solubility is as a key chemical and physical property. Solubility prediction methods are applied in diverse fields including preparation synthesis and modifications of materials. To overcome the shortcomings of existing solubility prediction methods, taking the mass transfer of two-phase system as an example, a solubility prediction model based on the diffusion theory and hybrid artificial intelligence method was proposed in this paper. An improved double-population chaotic accelerated particle swarm optimization (APSO) algorithm combined diffusion theory was developed according to the particle evolution utilizing diffusion energy. The developed algorithm was applied in the training of parameters of the radial basis function artificial neural network and then a model for predicting solubility was developed. The experimental results of supercritical carbon dioxide solubility in 8 polymers were consistent with the predicted values by the model, indicating the high prediction accuracy. The average relative deviation, squared correlation coefficient, and root mean square error were respectively 0.0036, 0.9970, and 0.0152, displaying its higher comprehensive performance. The model may also be applied in other physicochemical fields.

DeepSol: a deep learning framework for sequence-based protein solubility prediction.

Protein solubility plays a vital role in pharmaceutical research and production yield. For a given protein, the extent of its solubility can represent the quality of its function, and is ultimately defined by its sequence. Thus, it is imperative to develop novel, highly accurate in silico sequence-based protein solubility predictors. In this work we propose, DeepSol, a novel Deep Learning-based protein solubility predictor. The backbone of our framework is a convolutional neural network that exploits k-mer structure and additional sequence and structural features extracted from the protein sequence. DeepSol outperformed all known sequence-based state-of-the-art solubility prediction methods and attained an accuracy of 0.77 and Matthew's correlation coefficient of 0.55. The superior prediction accuracy of DeepSol allows to screen for sequences with enhanced production capacity and can more reliably predict solubility of novel proteins. DeepSol's best performing models and results are publicly deposited at https://doi.org/10.5281/zenodo.1162886 (Khurana and Mall, 2018). Supplementary data are available at Bioinformatics online.

Utilization of Grid partitioning based Fuzzy inference system approach as a novel method to estimate solubility of hydrocarbons in carbon dioxide

ABSTRACTRecently due to increasing demand for energy and declination of oil reservoir the researchers have been encouraged to investigate the enhancement of oil recovery (EOR) approaches. One of popular and wide applicable processes in EOR is carbon dioxide injection which is attractive for researchers and industries due to environmentally aspects, good efficiency in displacement and low cost. The carbon dioxide injection causes the hydrocarbons extracted from crude oil so the solubility of hydrocarbon in carbon dioxide which is one of the critical parameters affects this phenomenon becomes interesting topic for researchers. In the present work Grid partitioning based Fuzzy inference system approach as a new method for prediction of solubility of hydrocarbons in carbon dioxide as function of temperature, pressure and carbon number of alkane was applied. To show the accuracy of the model the coefficients of determination were determined as 0.9902 and 0.9584 for training and testing phases respectively.

A universal segment approach for the prediction of the activity coefficient of complex pharmaceuticals in non-electrolyte solvents

A novel method for the prediction of the activity coefficient in pharmaceutical–solvent systems is introduced. The method infers that the concept of segment interactions applies not only to components, but to the functional groups that comprise the component; more specifically, the popular modified UNIFAC (Dortmund) [1] functional groups. In the present work, four basic segment interactions are considered, that include hydrophobic (dispersive), hydrophilic, as well as positive and negative polarity. The predictive model was applied to a set of structurally diverse, complex pharmaceuticals, and compared to popular qualitative solubility prediction methods such as NRTL-SAC Chen and Song [2] and the UNIFAC based methods. Furthermore, the Akaike Information Criterion [3] and Focused Information Criterion [4] were used to establish the relative quality of the solubility predictions of various predictive models, with the new model exhibiting favourable results. The temperature dependence provided by the new model was also investigated.

Correlating and predicting the solubilities of polycyclic aromatic hydrocarbons in supercritical fluids using the compressed gas model and the reference solubilities

The solubilities of some polycyclic aromatic hydrocarbons (PAHs) in supercritical fluids were correlated and predicted using the compressed gas model and the reference solubilities. By introducing the reference solubilities, the compressed gas models combined with Peng–Robinson equation of state, Carnahan–Starling–van der Waals hard sphere equation of state and van der Waals one-parameter mixing rules can yield the overall averages of the average absolute relative deviations (AARDs) of 7.45% and 8.16% in solubility correlation for PAHs in supercritical fluids, respectively, which are much less than the values obtained using the compressed gas model combined with the corresponding equation of state and the sublimation pressures of the solutes. By introducing the reference solubilities, the calculated solubilities are not sensitive to the binary interaction parameters and the solubilities of PAHs in the corresponding supercritical fluid can be predicted with a specific binary interaction parameter. When using reference solubilities in the compressed gas model, the two equations of state provide comparable average AARDs in solubility prediction for PAHs in supercritical fluids, which are 14.00% and 14.19%, respectively. The introduction of reference solubilities into the compressed gas model provides a feasible and simple solubility prediction method for the compounds in supercritical fluids without using their sublimation pressures. Moreover, combined with the reference solubilities and the compressed gas model, the hard sphere equation of state provides a more simple method to predict the solubilities of solutes in supercritical fluids with only their solid molar volumes.

Solubility of Anthracene in C1−C3 Alcohols from (298.2 to 333.2) K and Their Mixtures with 2-Propanone at 298.2 K

The anthracene solubility in methanol, ethanol, 1-propanol, and 2-propanol from (298.2 to 333.2 K) and also in binary solvent mixtures of 2-propanone + C1-C3 alcohols at 298.2 K was reported. The van’t Hoff equation was used for correlation of the solubility data in monosolvents at different temperatures. The solubility of anthracene in mixed solvents was predicted using previously developed quantitative structure-property relationships (QSPR) that required knowledge of the solubility data in monosolvents. The overall mean percentage deviation of the correlated solubilities in monosolvents at different temperatures was 4.8 %. The corresponding value for solubility prediction methods in solvent mixtures varied from (8.5 to 11.9) %. Using ab initio prediction methods, the overall mean percentage deviations varied from (14.7 to 15.8) %.

Predicting Aqueous Solubility: The Role of Crystallinity

This paper revisits the role of crystallinity in predicting the aqueous solubility of a wide variety of organic compounds. Box and Comer (Current Drug Metabolism, 2008, 9, 869-878) fitted solubility data for 86 drugs to an equation based solely on log P. The General Solubility Equation of Jain and Yalkowsky, which accounts for the crystal lattice energy, was applied to the same data set and gives more accurate solubility predictions. In this simple comparison between two solubility prediction methods, we show that log P alone is only half of the solution, and that there is a need to include the melting point when dealing with crystalline solutes.

Application of adaptive neuro-fuzzy inference system for solubility prediction of carbon dioxide in polymers

Solubility of carbon dioxide in poly(vinyl acetate) (PVAc), poly(2,6-dimethyl-1,4-phenylene ether) (PPO), polypropylene (PP), and high-density polyethylene (HDPE), poly(butylene succinate) (PBS), poly(butylene succinate-co-adipate) (PBSA) and polystyrene (PS) are modeled by adaptive neuro-fuzzy inference system (ANFIS) in wide range of pressure and temperature. The results obtained in this work indicate that ANFIS is effective method for prediction of solubility of carbon dioxide in polymers and have better accuracy and simplicity compared with the classical methods.

Predicting Experimental Properties of Proteins from Sequence by Machine Learning Techniques

Efficient target selection methods are an important prerequisite for increasing the success rate and reducing the cost of high-throughput structural genomics efforts. There is a high demand for sequence-based methods capable of predicting experimentally tractable proteins and filtering out potentially difficult targets at different stages of the structural genomic pipeline. Simple empirical rules based on anecdotal evidence are being increasingly superseded by rigorous machine-learning algorithms. Although the simplicity of less advanced methods makes them more human understandable, more sophisticated formalized algorithms possess superior classification power. The quickly growing corpus of experimental success and failure data gathered by structural genomics consortia creates a unique opportunity for retrospective data mining using machine learning techniques and results in increased quality of classifiers. For example, the current solubility prediction methods are reaching the accuracy of over 70%. Furthermore, automated feature selection leads to better insight into the nature of the correlation between amino acid sequence and experimental outcome. In this review we summarize methods for predicting experimental success in cloning, expression, soluble expression, purification and crystallization of proteins with a special focus on publicly available resources. We also describe experimental data repositories and machine learning techniques used for classification and feature selection.

In silico prediction of aqueous solubility, human plasma protein binding and volume of distribution of compounds from calculated pKa and AlogP98 values.

We have investigated whether three important ADME (absorption, distribution, metabolism, excretion) related properties (aqueous solubility, human plasma protein binding, and human volume of distribution at steady-state) can be predicted from chemical structure alone if only the predicted predominant ionisation state and lipophilicity (calculated logP [P = octanol-water partition coefficient]) are considered. A simple, fast method for the in silico prediction of aqueous solubility of predominantly uncharged compounds has been developed, while some potential is shown for the prediction of predominantly charged or zwitterionic compounds. Ten other known in silico prediction methods for aqueous solubility have also been evaluated. It has furthermore been demonstrated that the molecular weight (MW) profile of training sets for the development of aqueous solubility prediction methods can influence their predictive performance with regard to test sets of either matching or diverging profiles. The same property descriptors which have been found most relevant for the prediction of aqueous solubility have also proved useful for the prediction of human plasma protein binding and human volume of distribution at steady-state.

Design of the Synthetic Route for Helical Peptides. Synthesis and Solubility of Model Peptides Having a Helical Structure

Abstract Fragment condensations between amino- and carboxyl-components of hydrophobic decapeptides, which have a β-sheet structure in the solid state and are insoluble in DMF, NMP, and DMSO, were examined in a mixture of CH2Cl2 and TFE (4/1, v/v) using various coupling reagents. By the use of DCC and HOBt as coupling reagents, the reaction proceeded smoothly in moderate yield to give eicosapeptides. These have a stable helical structure in the solid state and are easily soluble in a variety of organic solvents. The hydrophobic eicosapeptides obtained were subjected to a successive coupling reaction in CH2Cl2 alone to give helical tetracontapeptides in high yield, which also have high solubility in various organic solvents of low polarity. The solubility of hydrophobic helical peptides presents a solubility feature of helical peptides which are obtained as peptide intermediates in protein synthesis. The synthetic strategy for helical peptides and proteins is discussed in connection with the solubility prediction method.

Synthesis and solubility properties of peptide fragments of human hemoglobin alpha-chain (123-136).

The solubility prediction method for protected peptides was successfully applied to relatively small peptide fragments of human hemoglobin alpha-chain (123-136) which contained various polar amino acid residues such as Asp(OBzl), Glu(OBzl), Lys(Z), Ser(Bzl), and Thr(Bzl). As reported previously for hydrophobic peptides and human proinsulin C-peptide fragments, solubility data indicated that the insolubility of protected peptides having a mean value of Pc value below 0.90 appeared to begin at the octa- or nonapeptide sequence level and that beta-sheet structure played an important role in the insolubility of peptides. When a peptide has a beta-sheet structure in the solid state, we can clearly determine the critical chain length for peptide insolubility, the solubility dependence on solvent properties, and the solubility independence of amino acid compositions of peptides.

Infrared Absorption Study of Peptide Fragments of Human Hemoglobin α-Chain (123–136) in the Solid State

Abstract In order to evaluate the influence of polar amino acid residues on peptide conformations, IR absorption spectroscopic analysis of peptide fragments of human hemoglobin α-chain (123–136) and Boc–Leun–OBzl (n=3–6,9, and 12) was performed in the solid state. The polar amino acid residues involved in the peptide fragments are in the following: Asp(OBzl), Lys(Z), Ser(Bzl), and Thr(Bzl). Regardless of the amino acid composition, all of the peptides equal to or larger than a pentapeptide level exhibited a high potential for the formation of a β-sheet structure in the solid state. Conformational behaviors of peptides having polar amino acid residues are similar to those of Boc–Leun–OBzl, indicating that polar side chains can not cause disturbance of the formation of β-sheet structures. The fine relationship among the conformation, solubility, and 〈Pc〉 value of the peptide fragments strongly supports the propriety of the solubility prediction method for peptide intermediates. The influence of shear stress on peptide conformations in the solid state was also investigated precisely, and it was suggested that a heptapeptide size had the potential for the development of a helical structure in the solid state.

ChemInform Abstract: Design of the Synthetic Route for Peptides and Proteins Based on the Solubility Prediction Method. Part 1. Synthesis and Solubility Properties of Human Proinsulin C‐Peptide Fragments.

AbstractThe usefulness of the solubility prediction method is demonstrated using relatively small peptide fragments (I)‐(VII) of human proinsulin C‐peptide (VIII).

Design of the Synthetic Route for Peptides and Proteins Based on the Solubility Prediction Method. I. Synthesis and Solubility Properties of Human Proinsulin C-Peptide Fragments

Abstract The usefulness of the solubility prediction method is demonstrated using relatively small peptide fragments of human proinsulin C-peptide. The propriety of the solubility prediction method for peptides having polar side chains is also examined in the following respects: (1) Peptide intermediates smaller than a heptapeptide have high solubility regardless of their <Pc> values; (2) the <Pc> values of peptide intermediates are useful for judging solubility of peptide intermediates equal to or larger than an octapeptide level; (3) the Pro residue in a central position of a peptide chain is effective for increasing peptide solubility; and (4) there is critical chain length for peptide insolubility caused by a β-sheet aggregation. A strategy suitable for the design of the synthetic route for human proinsulin C-peptide is subsequently discussed on the basis of the solubility prediction of peptide intermediates.

Infrared Absorption Study of Human Proinsulin C-Peptide Fragments in the Solid State

Abstract In order to evaluate the propriety of the solubility prediction method for peptide intermediates having polar side chains, IR absorption spectroscopic analysis of human proinsulin C-peptide fragments was performed in the solid state. The polar amino acid residues involved in the peptide fragments are in the following: Glu(OBzl), Asp(OBzl), Gin, and Ser(Bzl). All peptide fragments except for ones containing the Pro residue exhibit a high potential for the formation of a β-sheet structure, indicating that, even when the polar side chains of the Glu, Asp, and Ser residues are protected by the Bzl group, the coil conformation parameter Pc for each amino acid residue is useful for estimating the potential for the formation of a β-sheet structure of peptide fragments smaller than critical size for development of α-helical structure in the solid state. As in the case of hydrophobic peptides, the β-sheet aggregation clearly plays an important role in reducing solubility of the peptide fragments having polar side chains. The effect of the Pro residue on the disturbance of the β-sheet structure was also observed. The ability of the short-range (polar side chain-backbone) or intraresidue interactions due to the low flexibility of the pyrrolidine ring of the Pro residue to promote helical folding in peptide fragments in the solid state was not observed for the peptide fragments corresponding to helical regions of human proinsulin.

Generalized Correlations for Predicting Solubility, Swelling and Viscosity Behavior of CO2 -Crude Oil Systems

Abstract This paper presents correlations for predicting the solubility, swelling and viscosity behavior of CO2-crude oil systems. The correlations were developed from experimental data obtained by the authors. These data are also presented. The data were determined by measuring the properties of mixtures of CO2 and nine different oils. Experimental conditions covered a range of 100 to 250 degrees F and pressures up to 2,300 psia. Properties predicted by the correlations have average deviations, expressed as per cent of experimental value, of 2 per cent for solubility, 0.5 per cent for swelling and 12 per cent for viscosity. Introduction Interest in CO2 injection as an oil recovery process has led to the development of performance prediction methods which can be applied to specific reservoirs. To use these performance prediction methods, it is necessary to know the solubility, swelling, and viscosity properties of CO2-crude oil mixtures at reservoir conditions. Some information on these properties has appeared in the literature; however, this information did not cover the range of different oils and conditions needed to prepare generalized correlations for reservoir engineering purposes. Consequently, an experimental program was undertaken to collect the data needed. The data obtained and the correlations developed from the data are described in the following sections of this paper. SOLUBILITY OF CO2 IN CRUDE OILS CO2 solubility data in the literature come from six principal sources. The solubility prediction method of Welker and Dunlop is limited to 80F.The information in Ref. 4 is of two types: the first includes binary and ternary mixtures of CO2 and light hydrocarbons (C1 to C6), and the second gives data for CO2 and heavy hydrocarbons for a temperature range of 40 to 90F.Ref. 5 contains a KCO2 chart for systems whose convergence pressure is 4,000 psia. The KCO2's are based mainly on CO2-natural gas mixtures. Poettmann's work covered CO2 solubility in one condensate and one crude oil. Jacoby and Rzasa measured CO2 solubilities as a function of pressure and temperature for two natural gas-absorber oil mixtures and two natural gas-crude oil mixtures. CO2 concentration in these four systems was fixed at 5 mol per cent. The work reported in this paper extends CO2 solubility data to a variety of different crude oil types in a temperature range from 110 to 250F and pressures up to 2,300 psia. The experimental procedure used by the authors to obtain the solubility data consisted of combining known amounts of pure CO2 and crude oil in a visual cell at a fixed temperature and measuring the bubble point of the mixture. Measurements were made for a total of 40 different CO2-oil mixtures and the results are shown in Table 2. The mixtures included nine different oils (seven crude oils and two refined oils) whose properties are listed in Table 1. All nine oils had vapor pressures less than 1 atm at the experimental temperatures. Consequently, analysis of the bubble-point vapor showed a CO2 concentration over 99 mol per cent. At no time during these experiments was a second, more dense, liquid phase observed. The solubility correlation which was developed from the data in Table 2 is presented in Figs. 1, 2 and 3. In these figures, solubility is expressed as xCO2, the mol fraction of CO2 in the CO2-Oil mixture. Fig. 1 shows solubility as a function of CO2 fugacity and temperature. Fig. 2 shows the same solubility data expressed as a function of saturation pressure and temperature. The solubility shown in Figs. 1 and 2 is for an oil whose UOP characterization factor is 11.7. UOP characterization factors of crude oils can be determined from Ref. 10 if the viscosity and API gravity of the oil are known. Fig. 3 gives the solubility correction factor for oils whose UOP characterization factors differ from 11.7. JPT P. 102ˆ