Comparison of approaches for rational siRNA design leading to a new efficient and transparent method

Current literature describes several methods for the design of efficient siRNAs with 19 perfectly matched base pairs and 2nt overhangs. Using four independent databases totaling 3336 experimentally verified siRNAs, we compared how well several of these methods predict siRNA cleavage efficiency. According to receiver operating characteristics (ROC) and correlation analyses, the best programs were BioPredsi, ThermoComposition and DSIR. We also studied individual parameters that significantly and consistently correlated with siRNA efficacy in different databases. As a result of this work we developed a new method which utilizes linear regression fitting with local duplex stability, nucleotide position-dependent preferences and total G/C content of siRNA duplexes as input parameters. The new method's discrimination ability of efficient and inefficient siRNAs is comparable with that of the best methods identified, but its parameters are more obviously related to the mechanisms of siRNA action in comparison with BioPredsi. This permits insight to the underlying physical features and relative importance of the parameters. The new method of predicting siRNA efficiency is faster than that of ThermoComposition because it does not employ time-consuming RNA secondary structure calculations and has much less parameters than DSIR. It is available as a web tool called ‘siRNA scales’.     

Requirement for canonical base pairing in the short pseudoknot structure of genomic hepatitis delta virus ribozyme

The tertiary structure of the 3′-cleaved product of the genomic hepatitis delta virus (HDV) ribozyme was solved by X-ray crystallographic analysis. In this structure, three single-stranded regions (SSrA, -B and -C) interact intricately with one another via hydrogen bonds between nucleotide bases, phosphate oxygens and 2′-OHs to form a nested double pseudoknot structure. Among these interactions, two Watson–Crick (W–C) base pairs, 726G–710C and 727G–709C, that form between SSrA and SSrC (P1.1) seem to be especially important for compact folding. To characterize the importance of these base pairs, ribozymes were subjected to in vitro selection from a pool of RNA molecules randomly substituted at positions 709, 710, 726 and 727. The results establish the importance of the two W–C base pairs for activity, although some mutants are active with one G–C base pair. In addition, the kinetic parameters were analyzed in all 16 combinations with two canonical base pairs. Comparison of variant ribozymes with the wild-type ribozyme reveals that the difference in reaction rates for these variants (ΔΔG^‡) is not simply accounted for by the differences in the stability of P1.1 (ΔΔG⁰₃₇). The role played by Mg²⁺ ions in formation of the P1.1 structure is also discussed.     

Energetic signatures of single base bulges: thermodynamic consequences and biological implications

DNA bulges are biologically consequential defects that can arise from template-primer misalignments during replication and pose challenges to the cellular DNA repair machinery. Calorimetric and spectroscopic characterizations of defect-containing duplexes reveal systematic patterns of sequence-context dependent bulge-induced destabilizations. These distinguishing energetic signatures are manifest in three coupled characteristics, namely: the magnitude of the bulge-induced duplex destabilization (ΔΔG_Bulge); the thermodynamic origins of ΔΔG_Bulge (i.e. enthalpic versus entropic); and, the cooperativity of the duplex melting transition (i.e. two-state versus non-two state). We find moderately destabilized duplexes undergo two-state dissociation and exhibit ΔΔG_Bulge values consistent with localized, nearest neighbor perturbations arising from unfavorable entropic contributions. Conversely, strongly destabilized duplexes melt in a non-two-state manner and exhibit ΔΔG_Bulge values consistent with perturbations exceeding nearest-neighbor expectations that are enthalpic in origin. Significantly, our data reveal an intriguing correlation in which the energetic impact of a single bulge base centered in one strand portends the impact of the corresponding complementary bulge base embedded in the opposite strand. We discuss potential correlations between these bulge-specific differential energetic profiles and their overall biological implications in terms of DNA recognition, repair and replication.     

The siRNA sequence and guide strand overhangs are determinants of in vivo duration of silencing

The use of short interfering RNAs (siRNA) in animals for target validation or as potential therapeutics is hindered by the short physical half-life when delivered as unencapsulated material and in turn the short active half-life of siRNAs in vivo. Here we demonstrate that the character of the two 3′-overhang nucleotides of the guide strand of siRNAs is a determinant of the duration of silencing by siRNAs both in vivo and in tissue culture cells. We demonstrate that deoxyribonucleotides in the guide strand overhang of siRNAs have a negative impact on maintenance of both the in vitro and in vivo activity of siRNAs over time. Overhangs that contain ribonucleotides or 2′-O-methyl modified nucleotides do not demonstrate this same impairment. We also demonstrate that the sequence of an siRNA is a determinant of the duration of silencing of siRNAs directed against the same target even when those siRNAs have equivalent activities in vitro. Our experiments have determined that a measurable duration parameter exists, distinct from both maximum silencing ability and the potency of siRNAs. Our findings provide information on incorporating chemically modified nucleotides into siRNAs for potent, durable therapeutics and also inform on methods used to select siRNAs for therapeutic and research purposes.     

Differing conformational pathways before and after chemistry for insertion of dATP versus dCTP opposite 8-oxoG in DNA polymerase beta

To elucidate how human DNA polymerase β (pol β) discriminates dATP from dCTP when processing 8-oxoguanine (8-oxoG), we analyze a series of dynamics simulations before and after the chemical step with dATP and dCTP opposite an 8-oxoG template started from partially open complexes of pol β. Analyses reveal that the thumb closing of pol β before chemistry is hampered when the incorrect nucleotide dATP is bound opposite 8-oxoG; the unfavorable interaction between active-site residue Tyr²⁷¹ and dATP that causes an anti to syn change in the 8-oxoG (syn):dATP complex explains this slow motion, in contrast to the 8-oxoG (anti):dCTP system. Such differences in conformational pathways before chemistry for mismatched versus matched complexes help explain the preference for correct insertion across 8-oxoG by pol β. Together with reference studies with a nonlesioned G template, we propose that 8-oxoG leads to lower efficiency in pol β's incorporation of dCTP compared with G by affecting the requisite active-site geometry for the chemical reaction before chemistry. Furthermore, because the active site is far from ready for the chemical reaction after partial closing or even full thumb closing, we suggest that pol β is tightly controlled not only by the chemical step but also by a closely related requirement for subtle active-site rearrangements after thumb movement but before chemistry.     

Bacillus subtilis polynucleotide phosphorylase 3′-to-5′ DNase activity is involved in DNA repair

In the presence of Mn²⁺, an activity in a preparation of purified Bacillus subtilis RecN degrades single-stranded (ss) DNA with a 3′ → 5′ polarity. This activity is not associated with RecN itself, because RecN purified from cells lacking polynucleotide phosphorylase (PNPase) does not show the exonuclease activity. We show here that, in the presence of Mn²⁺ and low-level inorganic phosphate (P_i), PNPase degrades ssDNA. The limited end-processing of DNA is regulated by ATP and is inactive in the presence of Mg²⁺ or high-level P_i. In contrast, the RNase activity of PNPase requires Mg²⁺ and P_i, suggesting that PNPase degradation of RNA and ssDNA occur by mutually exclusive mechanisms. A null pnpA mutation (ΔpnpA) is not epistatic with ΔrecA, but is epistatic with ΔrecN and Δku, which by themselves are non-epistatic. The addA5, ΔrecO, ΔrecQ (ΔrecJ), ΔrecU and ΔrecG mutations (representative of different epistatic groups), in the context of ΔpnpA, demonstrate gain- or loss-of-function by inactivation of repair-by-recombination, depending on acute or chronic exposure to the damaging agent and the nature of the DNA lesion. Our data suggest that PNPase is involved in various nucleic acid metabolic pathways, and its limited ssDNA exonuclease activity plays an important role in RecA-dependent and RecA-independent repair pathways.     

Design of nucleic acid sequences for DNA computing based on a thermodynamic approach

We have developed an algorithm for designing multiple sequences of nucleic acids that have a uniform melting temperature between the sequence and its complement and that do not hybridize non-specifically with each other based on the minimum free energy (ΔG_min). Sequences that satisfy these constraints can be utilized in computations, various engineering applications such as microarrays, and nano-fabrications. Our algorithm is a random generate-and-test algorithm: it generates a candidate sequence randomly and tests whether the sequence satisfies the constraints. The novelty of our algorithm is that the filtering method uses a greedy search to calculate ΔG_min. This effectively excludes inappropriate sequences before ΔG_min is calculated, thereby reducing computation time drastically when compared with an algorithm without the filtering. Experimental results in silico showed the superiority of the greedy search over the traditional approach based on the hamming distance. In addition, experimental results in vitro demonstrated that the experimental free energy (ΔG_exp) of 126 sequences correlated well with ΔG_min (|R| = 0.90) than with the hamming distance (|R| = 0.80). These results validate the rationality of a thermodynamic approach. We implemented our algorithm in a graphic user interface-based program written in Java.     

Analysis of the 3-hydroxyl group in Drosophila siRNA function

Members of the RNA-dependent RNA polymerase (RdRP) gene family have been shown to be essential for dsRNA-mediated gene silencing based on genetic screens in a variety of organisms, including Caenorhabditis elegans, Arabidopsis, Neurospora, and Dictyostelium. A hallmark of this process is the formation of small 21- to 25-bp dsRNAs, termed siRNAs for small interfering RNAs, which are derived from the dsRNA that initiates gene silencing. We have developed methods to demonstrate that these siRNAs produced in Drosophila embryo extract can be uniformly incorporated into dsRNA in a template-specific manner that is subsequently degraded by RNase III-related enzyme activity to create a second generation of siRNAs. SiRNA function in dsRNA synthesis and mRNA degradation depends upon the integrity of the 3^′-hydroxyl of the siRNA, consistent with the interpretation that siRNAs serve as primers for RdRP activity in the formation of dsRNA. This process of siRNA incorporation into dsRNA followed by degradation and the formation of new siRNAs has been termed “degradative PCR” and the proposed mechanism is consistent with the genetic and biochemical data derived from studies in C. elegans, Arabidopsis, Drosophila, and Dictyostelium. The methods used to study the function of both natural and synthetic siRNAs in RNA interference in Drosophila embryo extracts are detailed. The importance of the 3^′-hydroxyl group for siRNA function and its incorporation into dsRNA is emphasized and the results support a model that places RNA-dependent RNA polymerase as a key mediator in the RNA interference mechanism in Drosophila.     

The measurement of the intrinsic alkaline Bohr effect of various human haemoglobins by isoelectric focusing

We have used isoelectric focusing to measure the differences between the pI values of various normal and mutant human haemoglobins when completely deoxygenated and when fully liganded with CO. It was assumed that the ΔpI(deox.–ox.) values might correspond quantitatively to the intrinsic alkaline Bohr effect, as most of the anionic cofactors of the haemoglobin molecule are `stripped' off during the electrophoretic process. In haemoglobins known to exhibit a normal Bohr coefficient (ΔlogP₅₀/ΔpH) in solutions, the ΔpI(deox.–ox.) values are lower the higher their respective pI(ox.) values. This indicates that for any particular haemoglobin the ΔpI(deox.–ox.) value accounts for the difference in surface charges at the pH of its pI value. This was confirmed by measuring, by the direct-titration technique, the difference in pH of deoxy and fully liganded haemoglobin A₀ (α₂β₂) solutions in conditions approximating those of the isoelectric focusing, i.e. at 5°C and very low concentration of KCl. The variation of the ΔpH(deox.–ox.) curve as a function of pH (ox.) was similar to the isoelectric-focusing curve relating the variation of ΔpI(deox.–ox.) versus pI(ox.) in various haemoglobins with Bohr factor identical with that of haemoglobin A₀. In haemoglobin A₀ the ΔpI(deox.–ox.) value is 0.17 pH unit, which corresponds to a difference of 1.20 positive charges between the oxy and deoxy states of the tetrameric haemoglobin. This value compares favourably with the values of the intrinsic Bohr effect estimated in back-titration experiments. The ΔpI(deox.–ox.) values of mutant or chemically modified haemoglobins carrying an abnormality at the N- or C-terminus of the α-chains are decreased by 30% compared with the ΔpI value measured in haemoglobin A₀. When the C-terminus of the β-chains is altered, as in Hb Nancy (α₂β^{Tyr-145→Asp}₂), we observed a 70% decrease in the ΔpI value compared with that measured in haemoglobin A₀. These values are in close agreement with the estimated respective roles of the two major Bohr groups, Val-1α and His-146β, at the origin of the intrinsic alkaline Bohr effect [Kilmartin, Fogg, Luzzana & Rossi-Bernardi (1973) J. Biol. Chem. 248, 7039–7043; Perutz, Kilmartin, Nishikura, Fogg, Butler & Rollema (1980) J. Mol. Biol. 138, 649–670]. In other mutant haemoglobins it is demonstrated also that the ΔpI(deox.–ox.) value may be decreased or even suppressed when the substitution affects residues involved in the stability of the tetramer. These results support the interpretation proposed by Perutz, Kilmartin, Nishikura, Fogg, Butler & Rollema [(1980), J. Mol. Biol. 138, 649–670] for the mechanism of the alkaline Bohr effect, and also indicate that the transition between the two quaternary configurations is a prerequisite for the full expression of the alkaline Bohr effect.     

Action of nicotine and analogs on acetylcholine receptors having mutations of transmitter-binding site residue αG153

A primary target for nicotine is the acetylcholine receptor channel (AChR). Some of the ability of nicotine to activate differentially AChR subtypes has been traced to a transmitter-binding site amino acid that is glycine in lower affinity and lysine in higher affinity AChRs. We studied the effects of mutations of this residue (αG153) in neuromuscular AChRs activated by nicotine and eight other agonists including nornicotine and anabasine. All of the mutations increased the unliganded gating equilibrium constant. The affinity of the resting receptor (K_d) and the net binding energy from the agonist for gating (ΔG_B) were estimated by cross-concentration fitting of single-channel currents. In all but one of the agonist/mutant combinations there was a moderate decrease in K_d and essentially no change in ΔG_B. The exceptional case was nicotine plus lysine, which showed a large, >8,000-fold decrease in K_d but no change in ΔG_B. The extraordinary specificity of this combination leads us to speculate that AChRs with a lysine at position αG153 may be exposed to a nicotine-like compound in vivo.     

A General Framework for Thermodynamically Consistent Parameterization and Efficient Sampling of Enzymatic Reactions

Kinetic models provide the means to understand and predict the dynamic behaviour of enzymes upon different perturbations. Despite their obvious advantages, classical parameterizations require large amounts of data to fit their parameters. Particularly, enzymes displaying complex reaction and regulatory (allosteric) mechanisms require a great number of parameters and are therefore often represented by approximate formulae, thereby facilitating the fitting but ignoring many real kinetic behaviours. Here, we show that full exploration of the plausible kinetic space for any enzyme can be achieved using sampling strategies provided a thermodynamically feasible parameterization is used. To this end, we developed a General Reaction Assembly and Sampling Platform (GRASP) capable of consistently parameterizing and sampling accurate kinetic models using minimal reference data. The former integrates the generalized MWC model and the elementary reaction formalism. By formulating the appropriate thermodynamic constraints, our framework enables parameterization of any oligomeric enzyme kinetics without sacrificing complexity or using simplifying assumptions. This thermodynamically safe parameterization relies on the definition of a reference state upon which feasible parameter sets can be efficiently sampled. Uniform sampling of the kinetics space enabled dissecting enzyme catalysis and revealing the impact of thermodynamics on reaction kinetics. Our analysis distinguished three reaction elasticity regions for common biochemical reactions: a steep linear region (0> ΔG_r >-2 kJ/mol), a transition region (-2> ΔG_r >-20 kJ/mol) and a constant elasticity region (ΔG_r <-20 kJ/mol). We also applied this framework to model more complex kinetic behaviours such as the monomeric cooperativity of the mammalian glucokinase and the ultrasensitive response of the phosphoenolpyruvate carboxylase of Escherichia coli. In both cases, our approach described appropriately not only the kinetic behaviour of these enzymes, but it also provided insights about the particular features underpinning the observed kinetics. Overall, this framework will enable systematic parameterization and sampling of enzymatic reactions.     

Interaction with Btn2p is required for localization of Rsglp: Btn2p-mediated changes in arginine uptake in Saccharomyces cerevisiae

Btn2p, a novel coiled-coil protein, is up-regulated in btn1Δ yeast strains, and this up-regulation is thought to contribute to maintaining a stable vacuolar pH in btn1Δ strains (D. A. Pearce, T. Ferea, S. A. Nosel, B. Das, and F. Sherman, Nat. Genet. 22:55-58, 1999). We now report that Btn2p interacts biochemically and functionally with Rsg1p, a down-regulator of the Can1p arginine and lysine permease. Rsg1p localizes to a distinct structure toward the cell periphery, and strains lacking Btn2p (btn2Δ strains) fail to correctly localize Rsg1p. btn2Δ strains, like rsg1Δ strains, are sensitive for growth in the presence of the arginine analog canavanine. Furthermore, btn2Δ strains, like rsg1Δ strains, demonstrate an elevated rate of uptake of [¹⁴C]arginine, which leads to increased intracellular levels of arginine. Overexpression of BTN2 results in a decreased rate of arginine uptake. Collectively, these results indicate that altered levels of Btn2p can modulate arginine uptake through localization of the Can1p-arginine permease regulatory protein, Rsg1p. Our original identification of Btn2p was that it is up-regulated in the btn1Δ strain which serves as a model for the lysosomal storage disorder Batten disease. Btn1p is a vacuolar/lysosomal membrane protein, and btn1Δ suppresses both the canavanine sensitivity and the elevated rate of uptake of arginine displayed by btn2Δ rsg1Δ strains. We conclude that Btn2p interacts with Rsg1p and modulates arginine uptake. Up-regulation of BTN2 expression in btn1Δ strains may facilitate either a direct or indirect effect on intracellular arginine levels.     

Use of Mycobacterium smegmatis Deficient in ADP-Ribosyltransferase as Surrogate for Mycobacterium tuberculosis in Drug Testing and Mutation Analysis

Rifampicin (Rif) is a first line drug used for tuberculosis treatment. However, the emergence of drug resistant strains has necessitated synthesis and testing of newer analogs of Rif. Mycobacterium smegmatis is often used as a surrogate for M. tuberculosis. However, the presence of an ADP ribosyltransferase (Arr) in M. smegmatis inactivates Rif, rendering it impractical for screening of Rif analogs or other compounds when used in conjunction with them (Rif/Rif analogs). Rifampicin is also used in studying the role of various DNA repair enzymes by analyzing mutations in RpoB (a subunit of RNA polymerase) causing Rif resistance. These analyses use high concentrations of Rif when M. smegmatis is used as model. Here, we have generated M. smegmatis strains by deleting arr (Δarr). The M. smegmatis Δarr strains show minimum inhibitory concentration (MIC) for Rif which is similar to that for M. tuberculosis. The MICs for isoniazid, pyrazinamide, ethambutol, ciprofloxacin and streptomycin were essentially unaltered for M. smegmatis Δarr. The growth profiles and mutation spectrum of Δarr and, Δarr combined with ΔudgB (udgB encodes a DNA repair enzyme that excises uracil) strains were similar to their counterparts wild-type for arr. However, the mutation spectrum of ΔfpgΔarr strain differed somewhat from that of the Δfpg strain (fpg encodes a DNA repair enzyme that excises 8-oxo-G). Our studies suggest M. smegmatis Δarr strain as an ideal model system in drug testing and mutation spectrum determination in DNA repair studies.     

Predicting loop–helix tertiary structural contacts in RNA pseudoknots

Tertiary interactions between loops and helical stems play critical roles in the biological function of many RNA pseudoknots. However, quantitative predictions for RNA tertiary interactions remain elusive. Here we report a statistical mechanical model for the prediction of noncanonical loop–stem base-pairing interactions in RNA pseudoknots. Central to the model is the evaluation of the conformational entropy for the pseudoknotted folds with defined loop–stem tertiary structural contacts. We develop an RNA virtual bond-based conformational model (Vfold model), which permits a rigorous computation of the conformational entropy for a given fold that contains loop–stem tertiary contacts. With the entropy parameters predicted from the Vfold model and the energy parameters for the tertiary contacts as inserted parameters, we can then predict the RNA folding thermodynamics, from which we can extract the tertiary contact thermodynamic parameters from theory–experimental comparisons. These comparisons reveal a contact enthalpy (ΔH) of −14 kcal/mol and a contact entropy (ΔS) of −38 cal/mol/K for a protonated C⁺•(G–C) base triple at pH 7.0, and (ΔH = −7 kcal/mol, ΔS = −19 cal/mol/K) for an unprotonated base triple. Tests of the model for a series of pseudoknots show good theory–experiment agreement. Based on the extracted energy parameters for the tertiary structural contacts, the model enables predictions for the structure, stability, and folding pathways for RNA pseudoknots with known or postulated loop–stem tertiary contacts from the nucleotide sequence alone.     

Innate immunity to yeast prions: Btn2p and Cur1p curing of the [URE3] prion is prevented by 60S ribosomal protein deficiency or ubiquitin/proteasome system overactivity

[URE3] is an amyloid-based prion of Ure2p, a negative regulator of poor nitrogen source catabolism in Saccharomyces cerevisiae. Overproduced Btn2p or its paralog Cur1p, in processes requiring Hsp42, cure the [URE3] prion. Btn2p cures by collecting Ure2p amyloid filaments at one place in the cell. We find that rpl4aΔ, rpl21aΔ, rpl21bΔ, rpl11bΔ, and rpl16bΔ (large ribosomal subunit proteins) or ubr2Δ (ubiquitin ligase targeting Rpn4p, an activator of proteasome genes) reduce curing by overproduced Btn2p or Cur1p. Impaired curing in ubr2Δ or rpl21bΔ is restored by an rpn4Δ mutation. No effect of rps14aΔ or rps30bΔ on curing was observed, indicating that 60S subunit deficiency specifically impairs curing. Levels of Hsp42p, Sis1p, or Btn3p are unchanged in rpl4aΔ, rpl21bΔ, or ubr2Δ mutants. Overproduction of Cur1p or Btn2p was enhanced in rpn4Δ and hsp42Δ mutants, lower in ubr2Δ strains, and restored to above wild-type levels in rpn4Δ ubr2Δ strains. As in the wild-type, Ure2N-GFP colocalizes with Btn2-RFP in rpl4aΔ, rpl21bΔ, or ubr2Δ strains, but not in hsp42Δ. Btn2p/Cur1p overproduction cures [URE3] variants with low seed number, but seed number is not increased in rpl4aΔ, rpl21bΔ or ubr2Δ mutants. Knockouts of genes required for the protein sorting function of Btn2p did not affect curing of [URE3], nor did inactivation of the Hsp104 prion-curing activity. Overactivity of the ubiquitin/proteasome system, resulting from 60S subunit deficiency or ubr2Δ, may impair Cur1p and Btn2p curing of [URE3] by degrading Cur1p, Btn2p or another component of these curing systems.     

Mitogen-activated protein kinase Hog1 mediates adaptation to G1 checkpoint arrest during arsenite and hyperosmotic stress

Cells slow down cell cycle progression in order to adapt to unfavorable stress conditions. Yeast (Saccharomyces cerevisiae) responds to osmotic stress by triggering G₁ and G₂ checkpoint delays that are dependent on the mitogen-activated protein kinase (MAPK) Hog1. The high-osmolarity glycerol (HOG) pathway is also activated by arsenite, and the hog1Δ mutant is highly sensitive to arsenite, partly due to increased arsenite influx into hog1Δ cells. Yeast cell cycle regulation in response to arsenite and the role of Hog1 in this process have not yet been analyzed. Here, we found that long-term exposure to arsenite led to transient G₁ and G₂ delays in wild-type cells, whereas cells that lack the HOG1 gene or are defective in Hog1 kinase activity displayed persistent G₁ cell cycle arrest. Elevated levels of intracellular arsenite and “cross talk” between the HOG and pheromone response pathways, observed in arsenite-treated hog1Δ cells, prolonged the G₁ delay but did not cause a persistent G₁ arrest. In contrast, deletion of the SIC1 gene encoding a cyclin-dependent kinase inhibitor fully suppressed the observed block of G₁ exit in hog1Δ cells. Moreover, the Sic1 protein was stabilized in arsenite-treated hog1Δ cells. Interestingly, Sic1-dependent persistent G₁ arrest was also observed in hog1Δ cells during hyperosmotic stress. Taken together, our data point to an important role of the Hog1 kinase in adaptation to stress-induced G₁ cell cycle arrest.     

Predicting Thermodynamic Properties of PBXTHs with New Quantum Topological Indexes

Novel group quantitative structure-property relationship (QSPR) models on the thermodynamic properties of PBXTHs were presented, by the multiple linear regression (MLR) analysis method. Four thermodynamic properties were studied: the entropy (S^θ), the standard enthalpy of formation (Δ_fH^θ), the standard Gibbs energy of formation (Δ_fG^θ), and the relative standard Gibbs energy of formation (Δ_RG^θ). The results by the formula indicate that the calculated and predicted data in this study are in good agreement with those in literature and the deviation is within the experimental errors. To validate the estimation reliability for internal samples and the predictive ability for other samples, leave-one-out (LOO) cross validation (CV) and external validation were performed, and the results show that the models are satisfactory.