Showing papers by "Chris Sander published in 1993"

Improved prediction of protein secondary structure by use of sequence profiles and neural networks

Abstract: The explosive accumulation of protein sequences in the wake of large-scale sequencing projects is in stark contrast to the much slower experimental determination of protein structures. Improved methods of structure prediction from the gene sequence alone are therefore needed. Here, we report a substantial increase in both the accuracy and quality of secondary-structure predictions, using a neural-network algorithm. The main improvements come from the use of multiple sequence alignments (better overall accuracy), from "balanced training" (better prediction of beta-strands), and from "structure context training" (better prediction of helix and strand lengths). This method, cross-validated on seven different test sets purged of sequence similarity to learning sets, achieves a three-state prediction accuracy of 69.7%, significantly better than previous methods. In addition, the predicted structures have a more realistic distribution of helix and strand segments. The predictions may be suitable for use in practice as a first estimate of the structural type of newly sequenced proteins.

Convergent Evolution of Similar Enzymatic Function on Different Protein Folds: The Hexokinase, Ribokinase, and Galactokinase Families of Sugar Kinases

Abstract: Kinases that catalyze phosphorylation of sugars, called here sugar kinases, can be divided into at least three distinct nonhomologous families. The first is the hexokinase family, which contains many prokaryotic and eukaryotic sugar kinases with diverse specificities, including a new member, rhamnokinase from Salmonella typhimurium. The three-dimensional structure of hexokinase is known and can be used to build models of functionally important regions of other kinases in this family. The second is the ribokinase family, of unknown three-dimensional structure, and comprises pro- and eukaryotic ribokinases, bacterial fructokinases, the minor 6-phosphofructokinase 2 from Escherichia coli, 6-phosphotagatokinase, 1-phosphofructokinase, and, possibly, inosine-guanosine kinase. The third family, also of unknown three-dimensional structure, contains several bacterial and yeast galactokinases and eukaryotic mevalonate and phosphomevalonate kinases and may have a substrate binding region in common with homoserine kinases. Each of the three families of sugar kinases appears to have a distinct three-dimensional fold, since conserved sequence patterns are strikingly different for the three families. Yet each catalyzes chemically equivalent reactions on similar or identical substrates. The enzymatic function of sugar phosphorylation appears to have evolved independently on the three distinct structural frameworks, by convergent evolution. In addition, evolutionary trees reveal that (1) fructokinase specificity has evolved independently in both the hexokinase and ribokinase families and (2) glucose specificity has evolved independently in different branches of the hexokinase family. These are examples of independent Darwinian adaptation of a structure to the same substrate at different evolutionary times. The flexible combination of active sites and three-dimensional folds observed in nature can be exploited by protein engineers in designing and optimizing enzymatic function.

Quality control of protein models : directional atomic contact analysis

Abstract: Branden & Jones state, in Nature: `Protein crystallography is an exacting trade, and the results may contain errors that are difficult to identify. It is the crystallographer's responsibility to make sure that incorrect protein structures do not reach the literature.' [Branden & Jones. (1990). Nature (London), 343, 687–689.] One of several available methods of checking structures for correctness is the evaluation of atomic contacts. From an initial hypothesis that atom-atom interactions are the primary determinant of protein folding, any protein model can be tested for proper packing by the calculation of a contact quality index. The index is a measure of the agreement between the distributions of atoms around each residue fragment in the model and equivalent distributions derived from the database of known structures solved at high resolution. The better the agreement, the higher the contact quality index. This empirical test, which is independent of X-ray data, is applied to a series of successively refined crystal structures. In all cases, the model known or expected to be better (the one with the lower R-factor) has a better contact quality index, indicating that this type of contact analysis can be used as an independent quality criterion during crystallographic refinement. Modelled proteins and predicted mutant structures can also be evaluated.

An effective solvation term based on atomic occupancies for use in protein simulations

Abstract: Several approaches to the treatment of solvent effects based on continuum models are reviewed and a new method based on occupied atomic volumes (occupancies) is proposed and tested. The new method describes protein-water interactions in terms of atomic solvation parameters, which represent the solvation free energy per unit of volume. These parameters were determined for six different atoms types, using experimental free energies of solvation. The method was implemented in the GROMOS and PRESTO molecular simulation program suites. Simulations with the solvation term require 20-50% more CPU time than the corresponding vacuum simulations and are approximately 20 times faster than explicit water simulations. The method and parameters were tested by carrying out 200 ps simulations of BPTI in water, in vacuo, and with the solvation term. The performance of the solvation term was assessed by comparing the structures and energies from the solvation simulations with the equivalent quantities derived from...

The HSSP Database of Protein Structure-Sequence Alignments

Abstract: HSSP is a derived database merging structural (3-D) and sequence (1-D) information. For each protein of known 3-D structure from the Protein Data Bank (PDB), the database has a multiple sequence alignment of all available homologues and a sequence profile characteristic of the family. The list of homologues is the result of a database search in SwissProt using a position-weighted dynamic programming method for sequence profile alignment (MaxHom). The database is updated frequently. The listed homologues are very likely to have the same 3-D structure as the PDB protein to which they have been aligned. As a result, the database is not only a database of aligned sequence families, but also a database of implied secondary and tertiary structures covering 29% of all SwissProt-stored sequences.

Molecular modelling of the Norrie disease protein predicts a cystine knot growth factor tertiary structure.

Abstract: The X–lined gene for Norrie disease, which is characterized by blindness, deafness and mental retardation has been cloned recently. This gene has been thought to code for a putative extracellular factor; its predicted amino acid sequence is homologous to the C–terminal domain of diverse extracellular proteins. Sequence pattern searches and three–dimensional modelling now suggest that the Norrie disease protein (NDP) has a tertiary structure similar to that of transforming growth factor β (TGFβ). Our model identifies NDP as a member of an emerging family of growth factors containing a cystine knot motif, with direct implications for the physiological role of NDP. The model also sheds light on sequence related domains such as the C–terminal domain of mucins and of von Willebrand factor.

Modeling of transmembrane seven helix bundles.

Abstract: Transmembrane seven helix bundles form a large family of membrane inserted receptors and are responsible for a wide range of biological functions. Experimental data suggest that their overall structure is similar to bacteriorhodopsin. We describe here a new approach for the modeling of transmembrane seven helix bundles based on statistically derived environmental preference parameters combined with experimentally determined features of the receptors. The method was used to create a model for the human beta 2-adrenoreceptor. This model is physically plausible, is in reasonable agreement with experimental data and may be helpful in planning new receptor engineering experiments.

Secondary structure prediction of all-helical proteins in two states.

Abstract: Can secondary structure prediction be improved by prediction rules that focus on a particular structural class of proteins? To help answer this question, we have assessed the accuracy of prediction for all-helical proteins, using two conceptually different method and two levels of description. An overall two-state single-residue accuracy of ∼80% can be obtained by a neural network, no matter whether it is trained on two states (helix and non-helix) or first trained on three states (helix, strand and loop) and then evaluated on two states. For four test proteins, this is similar to the accuracy obtained with inductive logic programming

Hypothesis Does the HIV Nef protein mimic the MHC

Abstract: The sequence of the HIV Nef protein has no significant homology to other proteins in the SwissProt database, and experimental data concerning its function are sparse and contradictory. Using a novel protein sequence comparison method, we find similarities between different Nef sequences and the a chain of human MHC class I proteins. The possible biologica impli~tions of this f&ding are discussed.

From Sequence Similarity to Structural Homology of Proteins

Abstract: Attempts to solve the protein folding problem by empirical methods are said to be limited by the size of the database of known protein three-dimensional structures. Here we point out that this database can be significantly increased by the use of sequence homology, based on the following observations. (1) The database of known sequences, currently at more than 15000 proteins, is two orders of magnitude larger than the database of known structures. (2) The currently most powerful method of predicting protein structures is model building by homology. (3) Structural homology can be inferred from the level of sequence similarity. (4) The threshold of sequence similarity sufficient for structural homology depends strongly on the length of the alignment.

The prediction and design of protein structures

Abstract: Prediction of three-dimensional protein structure from sequence alone is a classical problem of molecular biology. Progress with this problem has been slow over the last 20 years. Using evolutionary information in the form of sequence and structure alignments of related proteins opens up powerful new approaches that bring us closer to a solution. For example, prediction of secondary structure has now been advanced to the 70% three-state accuracy level using a neural network algorithm with multiple related sequences as input [1]. In another example, structure comparison of actin with its distant evolutionary cousins led to a database search pattern that identified a new class of bacterial ATPases, probably ancient relatives of actin [2]. These approaches work because mutational noise and disparate functional requirements are averaged out, leaving a clearer sequence signal for the three-dimensional fold.