PubMed Health⌕ Search

Biomedical subjects

M J Sternberg

Publications and source records attributed to M J Sternberg.

At least 55 records · Page 3Linked to original sources
PubMed1996

Structure-activity relationships derived by machine learning: the use of atoms and their bond connectivities to predict mutagenicity by inductive logic programming.

We present a general approach to forming structure-activity relationships (SARs). This approach is based on representing chemical structure by atoms and their bond connectivities in combination with the inductive logic programming (ILP) algorithm PROGOL. Existing SAR methods describe chemical structure by using attributes which are general properties of an object. It is not possible to map chemical structure directly to attribute-based descriptions, as such descriptions have no internal organization. A more natural and general way to describe chemical structure is to use a relational description, where the internal construction of the description maps that of the object described. Our atom and bond connectivities representation is a relational description. ILP algorithms can form SARs with relational descriptions. We have tested the relational approach by investigating the SARs of 230 aromatic and heteroaromatic nitro compounds. These compounds had been split previously into two subsets, 188 compounds that were amenable to regression and 42 that were not. For the 188 compounds, a SAR was found that was as accurate as the best statistical or neural network-generated SARs. The PROGOL SAR has the advantages that it did not need the use of any indicator variables handcrafted by an expert, and the generated rules were easily comprehensible. For the 42 compounds, PROGOL formed a SAR that was significantly (P < 0.025) more accurate than linear regression, quadratic regression, and back-propagation. This SAR is based on an automatically generated structural alert for mutagenicity.

Algorithms↗
PubMed1996

Nitric oxide-mediated suppression of T cell responses during Trypanosoma brucei infection: soluble trypanosome products and interferon-gamma are synergistic inducers of nitric oxide synthase.

African trypanosome infections result in lymphocyte unresponsiveness and anemia in the mammalian host. In murine infections, these effects are mediated by suppressor macrophages releasing nitric oxide (NO). We investigated the mechanism of activation of macrophages to produce NO during trypanosomiasis in vitro. A soluble component of trypanosome lysates induced NO synthesis in peritoneal macrophage cultures only when the macrophages were co-stimulated with interferon-gamma (IFN-gamma). The macrophage-activating factor was also released in a soluble form by live bloodstream-form trypanosomes, but not procyclic trypanosomes. When splenocyte cultures were exposed to IFN-gamma and trypanosomes, an NO-dependent suppression of T cell proliferation occurred. This is similar to the suppression observed in the spleens of trypanosome-infected mice, suggesting that a combination of trypanosome-released macrophage-activating factors and IFN-gamma are a trigger of immune dysfunction in trypanosomiasis.

Animals↗
PubMed1996

Identification and application of the concepts important for accurate and reliable protein secondary structure prediction.

A protein secondary structure prediction method from multiply aligned homologous sequences is presented with an overall per residue three-state accuracy of 70.1%. There are two aims: to obtain high accuracy by identification of a set of concepts important for prediction followed by use of linear statistics; and to provide insight into the folding process. The important concepts in secondary structure prediction are identified as: residue conformational propensities, sequence edge effects, moments of hydrophobicity, position of insertions and deletions in aligned homologous sequence, moments of conservation, auto-correlation, residue ratios, secondary structure feedback effects, and filtering. Explicit use of edge effects, moments of conservation, and auto-correlation are new to this paper. The relative importance of the concepts used in prediction was analyzed by stepwise addition of information and examination of weights in the discrimination function. The simple and explicit structure of the prediction allows the method to be reimplemented easily. The accuracy of a prediction is predictable a priori. This permits evaluation of the utility of the prediction: 10% of the chains predicted were identified correctly as having a mean accuracy of > 80%. Existing high-accuracy prediction methods are "black-box" predictors based on complex nonlinear statistics (e.g., neural networks in PHD: Rost & Sander, 1993a). For medium- to short-length chains (> or = 90 residues and < 170 residues), the prediction method is significantly more accurate (P < 0.01) than the PHD algorithm (probably the most commonly used algorithm). In combination with the PHD, an algorithm is formed that is significantly more accurate than either method, with an estimated overall three-state accuracy of 72.4%, the highest accuracy reported for any prediction method.

Amino Acid Sequence↗
PubMed1995

Evidence that aspartic acid 301 is a critical substrate-contact residue in the active site of cytochrome P450 2D6.

Model building studies have intimated a role for aspartic acid 301 in the substrate binding of cytochrome P450 2D6 (CYP2D6). We have tested this hypothesis by generating a range of CYP2D6 mutants substituting a variety of amino acids at this site. The mutant proteins, which included substitution with a negatively charged glutamic acid residue or neutral asparagine, alanine, or glycine residues, were expressed in Saccharomyces cerevisiae. In addition, a mutant where aspartic acid 301 was deleted was also tested. All the mutants expressed approximately equivalent amounts of recombinant apoprotein and, apart from the alanine 301 and the aspartic acid 301 deletion mutants, gave carbon monoxide difference spectra of similar magnitude to the wild type. In the cases of the alanine and deletion mutants, the amount of holoprotein was significantly reduced or absent relative to the amount of apoprotein, indicating restricted heme incorporation. The glutamic acid mutant was shown to have similar catalytic properties to the wild type enzyme toward the substrates debrisoquine and metoprolol; however, some differences in regioselectivity and ligand binding were observed. The mutants containing neutral amino acids at position 301 exhibited marked reductions in catalytic activity. At low substrate concentrations little, if any, activity toward debrisoquine and metoprolol was measured. However, at a higher substrate concentration (2 mM) some activity was observed (about 10-20% of wild type levels). Consistent with the above findings, the debrisoquine-induced spin changes in the mutant proteins were markedly reduced. These data collectively demonstrate that aspartic acid 301 plays an important role in determining the substrate specificity and activity of CYP2D6 and provide experimental evidence supporting the role of this amino acid in forming an electrostatic interaction between the basic nitrogen atom in CYP2D6 substrates and the carboxylate group of aspartic acid 301.

Aspartic Acid↗
PubMed1995

A continuum model for protein-protein interactions: application to the docking problem.

The prediction of protein-protein interactions in solution is a major goal of theoretical structural biology. Here, we implement a continuum description of the thermodynamic processes involved. The model differs considerably from previous models in its use of "molecular surface" area to describe the hydrophobic component to the free energy of conformational change in solution. We have applied this model to a data set of alternative docked conformations of protein-protein complexes which were generated independently of this work. It was found previously that commonly used energy evaluation techniques fail to distinguish between near-native and certain non-native complexes in this data set. Here, we found that an energy function that takes into account (1) total electrostatic free energy, (2) hydrophobic free energy and (3) loss in side-chain conformational energy was able to reliably discriminate between near-native and non-native configurations but only when molecular surface is used as a descriptor of the hydrophobic effect. It is shown that the molecular surface and the more conventional surface descriptor "solvent accessible surface" give very different quantitative measures of hydrophobicity. In terms of the contribution of different energy components to the free energy of complex formation it was found that loss in side-chain conformational entropy is a second order effect. Electrostatic interaction energy (which is commonly used to score docked conformations) was a poor indicator of complementarity when starting from unbound conformations. It was found that electrostatic desolvation energy and the hydrophobic contribution (based on a molecular surface area descriptor) are much less sensitive to local fluctuations in atomic structure than point-to-point interaction energies and thus may be more suited for use as a scoring function when docking unbound conformations, where atomic complementarity is much less apparent. Whilst a combined energy function was able to distinguish near-native from non-native conformations in the six systems studied here, it remains to be determined to what extent more sizeable conformational changes would influence the results.

Algorithms↗
PubMed1995

Structure prediction. How good are we?

Recent successes show that, in certain circumstances, protein secondary structures can be predicted with high accuracy. How far are we from being able to predict the complete structure of a protein from its sequence?

Amino Acid Sequence↗
PubMed1995

Side-chain conformational entropy in protein folding.

An important, but often neglected, contribution to the thermodynamics of protein folding is the loss of entropy that results from restricting the number of accessible side-chain conformers in the native structure. Conformational entropy changes can be found by comparing the number of accessible rotamers in the unfolded and folded states, or by estimating fusion entropies. Comparison of several sets of results using different techniques shows that the mean conformational free energy change (T delta S) is 1 kcal.mol-1 per side chain or 0.5 kcal.mol-1 per bond. Changes in vibrational entropy appear to be negligible compared to the entropy change resulting from the loss of accessible rotamers. Side-chain entropies can help rationalize alpha-helix propensities, predict protein/inhibitor complex structures, and account for the distribution of side chains on the protein surface or interior.

Peptide Fragments↗
PubMed1995

Identification and analysis of domains in proteins.

An automatic algorithm based on inter-residue contacts is presented to identify domains in proteins. The results of the algorithm are compared to an assignment performed by inspection that was guided by the authors' description in the literature. The authors' and the algorithm's assignments for a chain were considered to agree if the same number of domains were identified and if the assignments were the same for at least 95% of the residues. With this criterion, the algorithm agreed with the authors' assignment for 78% of the 284 non-redundant chains considered. When some of the authors' assignments were re-evaluated based on the results of the algorithm, an agreement of 84% was obtained. The algorithm is therefore a useful tool for data validation in domain assignment. The authors assignments of domains were analysed for structural principles of domains. The number of chains forming one, two, three, four and five domains are 197, 67, 13, 6 and 1 respectively. Most domains in multidomain proteins are formed from continuous segments and adopt the same structural class. Distributions of the number of residues and the ellipticity of domains and chains are presented. The relationship between accessible surface area and molecular weight for domains and chains is examined.

Algorithms↗
PubMed1995

Towards an intelligent system for the automatic assignment of domains in globular proteins.

The automatic identification of protein domains from coordinates is the first step in the classification of protein folds and hence is required for databases to guide structure prediction. Most algorithms encode a single concept based and sometimes do not yield assignments that are consistent with the generally accepted perception. Our development of an automatic approach to identify reliably domains from protein coordinates is described. The algorithm is benchmarked against a manual identification of the domains in 284 representative protein chains. The first step is the domain assignment by distance (DAD) algorithm that considers the density of inter-residue contacts represented in a contact matrix. The algorithm yields 85% agreement with the manual assignment. The paper then considers how the reliability of these assignments could be evaluated. Finally the use of structural comparisons using the STAMP algorithm to validate domain assignment is reported on a test case.

Algorithms↗
PubMed1994

Analysis and classification of disulphide connectivity in proteins. The entropic effect of cross-linkage.

An analysis and a classification of protein disulphide connectivity in a set of distinct sequences are presented. We analyse the number of disulphides per sequence, the number of disulphides per residue and the length of disulphide cross-linked loops. Observed connectivities are classified according to the different possible types of arrangement. In addition, we classify disulphide connectivity by physical models describing the arrangement of multiple disulphides. Firstly, we consider whether the features of native connectivity arrangement are describable by the likelihood of diffusive contact in the unfolded state. This is referred to as the diffusion model, and was originated by Kauzmann. A second model, effectively the inverse of the diffusion model, describes native connectivity arrangement as dominated by the entropic stabilisation effect of cross-linkage. This is referred to as the entropic model. Additionally, we compare the distribution of disulphide cross-linked loops and of loops formed by disulphide-like contacts. For short sequences (less than about 75 residues) native connectivities tend to have entropically more-stabilising arrangement features, whilst for longer sequences (greater than about 200 residues) the diffusion model is appropriate. We introduce the concept of arrangement entropy as a measure of the complexity of a connectivity.

Disulfides↗
PubMed1994

Application of machine learning to structural molecular biology.

A technique of machine learning, inductive logic programming implemented in the program GOLEM, has been applied to three problems in structural molecular biology. These problems are: the prediction of protein secondary structure; the identification of rules governing the arrangement of beta-sheets strands in the tertiary folding of proteins; and the modelling of a quantitative structure activity relationship (QSAR) of a series of drugs. For secondary structure prediction and the QSAR, GOLEM yielded predictions comparable with contemporary approaches including neural networks. Rules for beta-strand arrangement are derived and it is planned to contrast their accuracy with those obtained by human inspection. In all three studies GOLEM discovered rules that provided insight into the stereochemistry of the system. We conclude machine learning used together with human intervention will provide a powerful tool to discover patterns in biological sequences and structures.

Amino Acid Sequence↗
PubMed1994

Conservation of polyproline II helices in homologous proteins: implications for structure prediction by model building.

Left-handed polyproline II (PPII) helices commonly occur in globular proteins in segments of 4-8 residues. This paper analyzes the structural conservation of PPII-helices in 3 protein families: serine proteinases, aspartic proteinases, and immunoglobulin constant domains. Calculations of the number of conserved segments based on structural alignment of homologous molecules yielded similar results for the PPII-helices, the alpha-helices, and the beta-strands. The PPII-helices are consistently conserved at the level of 100-80% in the proteins with sequence identity above 20% and RMS deviation of structure alignments below 3.0 A. The most structurally important PPII segments are conserved below this level of sequence identity. These results suggest that the PPII-helices, in addition to the other 2 secondary structure classes, should be identified as part of structurally conserved regions in proteins. This is supported by similar values for the local RMS deviations of the aligned segments for the structural classes of PPII-helices, alpha-helices, and beta-strands. The PPII-helices are shown to participate in supersecondary elements such as PPII-helix/alpha-helix. The conservation of PPII-helices depends on the conservation of a supersecondary element as a whole. PPII-helices also form links, possibly flexible, in the interdomain regions. The role of the PPII-helices in model building by homology is 2-fold; they serve as additional conserved elements in the structure allowing improvement of the accuracy of a model and provide correct chain geometry for modeling of the segments equivalenced to them in a target sequence. The improvement in model building is demonstrated in 2 test studies.

Amino Acid Sequence↗
PubMed1994

Quantitative structure-activity relationships by neural networks and inductive logic programming. I. The inhibition of dihydrofolate reductase by pyrimidines.

Neural networks and inductive logic programming (ILP) have been compared to linear regression for modelling the QSAR of the inhibition of E. coli dihydrofolate reductase (DHFR) by 2,4-diamino-5-(substituted benzyl)pyrimidines, and, in the subsequent paper [Hirst, J.D., King, R.D. and Sternberg, M.J.E. J. Comput.-Aided Mol. Design, 8 (1994) 421], the inhibition of rodent DHFR by 2,4-diamino-6,6-dimethyl-5-phenyl-dihydrotriazines. Cross-validation trials provide a statistically rigorous assessment of the predictive capabilities of the methods, with training and testing data selected randomly and all the methods developed using identical training data. For the ILP analysis, molecules are represented by attributes other than Hansch parameters. Neural networks and ILP perform better than linear regression using the attribute representation, but the difference is not statistically significant. The major benefit from the ILP analysis is the formulation of understandable rules relating the activity of the inhibitors to their chemical structure.

Animals↗
PubMed1994

Quantitative structure-activity relationships by neural networks and inductive logic programming. II. The inhibition of dihydrofolate reductase by triazines.

One of the largest available data sets for developing a quantitative structure-activity relationship (QSAR)--the inhibition of dihydrofolate reductase (DHFR) by 2,4-diamino-6,6-dimethyl-5-phenyl-dihydrotriazine derivatives--has been used for a sixfold cross-validation trial of neural networks, inductive logic programming (ILP) and linear regression. No statistically significant difference was found between the predictive capabilities of the methods. However, the representation of molecules by attributes, which is integral to the ILP approach, provides understandable rules about drug-receptor interactions.

Animals↗
PubMed1994

On the use of machine learning to identify topological rules in the packing of beta-strands.

The machine learning program GOLEM was applied to discover topological rules in the packing of beta-sheets in alpha/beta-domain proteins. Rules (constraints) were determined for four features of beta-sheet packing: (i) whether a beta-strand is at an edge; (ii) whether two consecutive beta-strands pack parallel or anti-parallel; (iii) whether two beta-strands pack adjacently; and (iv) the winding direction of two consecutive beta-strands. Rules were found with high predictive accuracy and coverage. The errors were generally associated with complications in domain folds, especially in one doubly would domains. Investigation of the rules revealed interesting patterns, some of which were known previously, others that are novel. Novel features include (i) the relationship between pairs of sequential strands is in general one of decreasing size; (ii) more sequential pairs of strands wind in the direction out than in; and (iii) it takes a larger alteration in hydrophobicity to change a strand from winding in the direction out than in. These patterns in the data may be the result of folding pathways in the domains. The rules found are of predictive value and could be used in the combinatorial prediction of protein structure, or as a general test of model structures, e.g. those produced by threading. We conclude that machine learning has a useful role in the analysis of protein structures.

Amino Acid Sequence↗
PubMed1994

Protein side-chain conformational entropy derived from fusion data--comparison with other empirical scales.

The loss of conformational entropy of protein side-chains is a major effect in the energetics of folding. The simplest approach is to enumerate the number of freely rotatable bonds. Recently, two scales of side-chain conformational entropy have been proposed based on the definition of entropy as the Boltzmann sampling over all accessible states (S = -R sigma p(i)lnp(i), where p(i) is the probability of being in a rotameric state). In one scale, derived only for aliphatic and aromatic side-chains, the values of p(i) were obtained from Monte Carlo simulations. In the other scale, the observed frequencies of different rotameric states in a database of protein crystal structures yielded an estimate for p(i). Here an empirical estimation of the fusion entropy of the side-chains is used to derive a third scale. The fusion entropy is obtained as a sum of empirically derived contributions from component hydrocarbon and functional groups. There is a good agreement between the fusion scale and the other two scales. This suggests that the magnitude of conformational entropy is being correctly established.

Computer Simulation↗