Burkhard Rost

The PredictProtein server

PredictProtein (http://www.predictprotein.org) is an Internet service for sequence analysis and the prediction of protein structure and function. Users submit protein sequences or alignments; PredictProtein returns multiple sequence alignments, PROSITE sequence motifs, low-complexity regions (SEG), nuclear localization signals, regions lacking regular structure (NORS) and predictions of secondary structure, solvent accessibility, globular regions, transmembrane helices, coiled-coil regions, structural switch regions, disulfide-bonds, sub-cellular localization and functional annotations. Upon request fold recognition by prediction-based threading, CHOP domain assignments, predictions of transmembrane strands and inter-residue contacts are also available. For all services, users can submit their query either by electronic mail or interactively via the World Wide Web.

Improving the prediction of protein secondary structure in three and eight classes using recurrent neural networks and profiles.

Secondary structure predictions are increasingly becoming the workhorse for several methods aiming at predicting protein structure and function. Here we use ensembles of bidirectional recurrent neural network architectures, PSI‐BLAST‐derived profiles, and a large nonredundant training set to derive two new predictors: (a) the second version of the SSpro program for secondary structure classification into three categories and (b) the first version of the SSpro8 program for secondary structure classification into the eight classes produced by the DSSP program. We describe the results of three different test sets on which SSpro achieved a sustained performance of about 78% correct prediction. We report confusion matrices, compare PSI‐BLAST to BLAST‐derived profiles, and assess the corresponding performance improvements. SSpro and SSpro8 are implemented as web servers, available together with other structural feature predictors at: http://promoter.ics.uci.edu/BRNN‐PRED/. Proteins 2002;47:228–235.

Finding nuclear localization signals

A variety of nuclear localization signals (NLSs) are experimentally known although only one motif was available for database searches through PROSITE. We initially collected a set of 91 experimentally verified NLSs from the literature. Through iterated ‘in silico mutagenesis’ we then extended the set to 214 potential NLSs. This final set matched in 43% of all known nuclear proteins and in no known non‐nuclear protein. We estimated that >17% of all eukaryotic proteins may be imported into the nucleus. Finally, we found an overlap between the NLS and DNA‐binding region for 90% of the proteins for which both the NLS and DNA‐binding regions were known. Thus, evolution seemed to have used part of the existing DNA‐binding mechanism when compartmentalizing DNA‐binding proteins into the nucleus. However, only 56 of our 214 NLS motifs overlapped with DNA‐binding regions. These 56 NLSs enabled a de novo prediction of partial DNA‐binding regions for ∼800 proteins in human, fly, worm and yeast.

PHD-an automatic mail server for protein secondary structure prediction

By the middle of 1993, > 30,000 protein sequences has been listed. For 1000 of these, the three-dimensional (tertiary) structure has been experimentally solved. Another 7000 can be modelled by homology. For the remaining 21,000 sequences, secondary structure prediction provides a rough estimate of structural features. Predictions in three states range between 35% (random) and 88% (homology modelling) overall accuracy. Using information about evolutionary conservation as contained in multiple sequence alignments, the secondary structure of 4700 protein sequences was predicted by the automatic e-mail server PHD. For proteins with at least one known homologue, the method has an expected overall three-state accuracy of 71.4% for proteins with at least one known homologue (evaluated on 126 unique protein chains).

SNAP: predict effect of non-synonymous polymorphisms on function

Many genetic variations are single nucleotide polymorphisms (SNPs). Non-synonymous SNPs are ‘neutral’ if the resulting point-mutated protein is not functionally discernible from the wild type and ‘non-neutral’ otherwise. The ability to identify non-neutral substitutions could significantly aid targeting disease causing detrimental mutations, as well as SNPs that increase the fitness of particular phenotypes. Here, we introduced comprehensive data sets to assess the performance of methods that predict SNP effects. Along we introduced SNAP (screening for non-acceptable polymorphisms), a neural network-based method for the prediction of the functional effects of non-synonymous SNPs. SNAP needs only sequence information as input, but benefits from functional and structural annotations, if available. In a cross-validation test on over 80 000 mutants, SNAP identified 80% of the non-neutral substitutions at 77% accuracy and 76% of the neutral substitutions at 80% accuracy. This constituted an important improvement over other methods; the improvement rose to over ten percentage points for mutants for which existing methods disagreed. Possibly even more importantly SNAP introduced a well-calibrated measure for the reliability of each prediction. This measure will allow users to focus on the most accurate predictions and/or the most severe effects. Available at http://www.rostlab.org/services/SNAP

A mutation in VPS35, encoding a subunit of the retromer complex, causes late-onset Parkinson disease.

To identify rare causal variants in late-onset Parkinson disease (PD), we investigated an Austrian family with 16 affected individuals by exome sequencing. We found a missense mutation, c.1858G>A (p.Asp620Asn), in the VPS35 gene in all seven affected family members who are alive. By screening additional PD cases, we saw the same variant cosegregating with the disease in an autosomal-dominant mode with high but incomplete penetrance in two further families with five and ten affected members, respectively. The mean age of onset in the affected individuals was 53 years. Genotyping showed that the shared haplotype extends across 65 kilobases around VPS35. Screening the entire VPS35 coding sequence in an additional 860 cases and 1014 controls revealed six further nonsynonymous missense variants. Three were only present in cases, two were only present in controls, and one was present in cases and controls. The familial mutation p.Asp620Asn and a further variant, c.1570C>T (p.Arg524Trp), detected in a sporadic PD case were predicted to be damaging by sequence-based and molecular-dynamics analyses. VPS35 is a component of the retromer complex and mediates retrograde transport between endosomes and the trans-Golgi network, and it has recently been found to be involved in Alzheimer disease.

Three-Dimensional Structures of Membrane Proteins from Genomic Sequencing

We show that amino acid covariation in proteins, extracted from the evolutionary sequence record, can be used to fold transmembrane proteins. We use this technique to predict previously unknown 3D structures for 11 transmembrane proteins (with up to 14 helices) from their sequences alone. The prediction method (EVfold_membrane) applies a maximum entropy approach to infer evolutionary covariation in pairs of sequence positions within a protein family and then generates all-atom models with the derived pairwise distance constraints. We benchmark the approach with blinded de novo computation of known transmembrane protein structures from 23 families, demonstrating unprecedented accuracy of the method for large transmembrane proteins. We show how the method can predict oligomerization, functional sites, and conformational changes in transmembrane proteins. With the rapid rise in large-scale sequencing, more accurate and more comprehensive information on evolutionary constraints can be decoded from genetic variation, greatly expanding the repertoire of transmembrane proteins amenable to modeling by this method.

Enzyme Function Less Conserved than Anticipated

The level of sequence similarity that implies similarity in protein structure is well established. Recently, many groups proposed thresholds for similarity in sequence implying similarity in enzymatic function. All previous results suggest the strong conservation of enzymatic function above levels of 50% pairwise sequence identity. Here, I argue that all groups substantially overestimated the conservation of enzyme function because their data sets were either too biased, or too small. An unbiased analysis suggested that less than 30% of the pair fragments above 50% sequence identity have entirely identical EC numbers. Another surprising finding was that even BLAST E-values below 10(-50) did not suffice to automatically transfer enzyme function without errors. As expected, most misclassifications originated from similarities in relatively short regions and/or from transferring annotations for different domains. Both problems cannot be corrected easily by adjusting the thresholds for automatic transfer of genome annotations. A score relating sequence identity to alignment length (distance from HSSP-threshold) outperformed statistical BLAST scores for high sequence similarity. In particular, the distance score allowed error-free transfer of enzyme function for the 10% most similar enzyme pairs. The results illustrated how difficult it is to assess the conservation of protein function and to guarantee error-free genome annotations, in general: sets with millions of pair comparisons might not suffice to arrive at statistically significant conclusions. In practice, the revised detailed estimates for the sequence conservation of enzyme function may provide important benchmarks for everyday sequence analysis and for more cautious automatic genome annotations.

Analysing six types of protein-protein interfaces.

Non-covalent residue side-chain interactions occur in many different types of proteins and facilitate many biological functions. Are these differences manifested in the sequence compositions and/or the residue-residue contact preferences of the interfaces? Previous studies analysed small data sets and gave contradictory answers. Here, we introduced a new data-mining method that yielded the largest high-resolution data set of interactions analysed. We introduced an information theory-based analysis method. On the basis of sequence features, we were able to differentiate six types of protein interfaces, each corresponding to a different functional or structural association between residues. Particularly, we found significant differences in amino acid composition and residue-residue preferences between interactions of residues within the same structural domain and between different domains, between permanent and transient interfaces, and between interactions associating homo-oligomers and hetero-oligomers. The differences between the six types were so substantial that, using amino acid composition alone, we could predict statistically to which of the six types of interfaces a pool of 1000 residues belongs at 63-100% accuracy. All interfaces differed significantly from the background of all residues in SWISS-PROT, from the group of surface residues, and from internal residues that were not involved in non-trivial interactions. Overall, our results suggest that the interface type could be predicted from sequence and that interface-type specific mean-field potentials may be adequate for certain applications.

A modified definition of Sov, a segment‐based measure for protein secondary structure prediction assessment

We present a measure for the evaluation of secondary structure prediction methods that is based on secondary structure segments rather than individual residues. The algorithm is an extension of the segment overlap measure Sov, originally defined by Rost et al. (J Mol Biol 1994;235:13–26). The new definition of Sov corrects the normalization procedure and improves Sovs ability to discriminate between similar and dissimilar segment distributions. The method has been comprehensively tested during the second Critical Assessment of Techniques for Protein Structure Prediction (CASP2). Here, we describe the underlying concepts, modifications to the original definition, and their significance. Proteins 1999;34:220–223. Published 1999 Wiley‐Liss, Inc.