Oliviero Carugo

Detailed estimation of bioinformatics prediction reliability through the Fragmented Prediction Performance Plots

BackgroundAn important and yet rather neglected question related to bioinformatics predictions is the estimation of the amount of data that is needed to allow reliable predictions. Bioinformatics predictions are usually validated through a series of figures of merit, like for example sensitivity and precision, and little attention is paid to the fact that their performance may depend on the amount of data used to make the predictions themselves.ResultsHere I describe a tool, named Fragmented Prediction Performance Plot (FPPP), which monitors the relationship between the prediction reliability and the amount of information underling the prediction themselves. Three examples of FPPPs are presented to illustrate their principal features. In one example, the reliability becomes independent, over a certain threshold, of the amount of data used to predict protein features and the intrinsic reliability of the predictor can be estimated. In the other two cases, on the contrary, the reliability strongly depends on the amount of data used to make the predictions and, thus, the intrinsic reliability of the two predictors cannot be determined. Only in the first example it is thus possible to fully quantify the prediction performance.ConclusionIt is thus highly advisable to use FPPPs to determine the performance of any new bioinformatics prediction protocol, in order to fully quantify its prediction power and to allow comparisons between two or more predictors based on different types of data.

Direct interaction of actin filaments with F‐BAR protein pacsin2

Two mechanisms have emerged as major regulators of membrane shape: BAR domain‐containing proteins, which induce invaginations and protrusions, and nuclear promoting factors, which cause generation of branched actin filaments that exert mechanical forces on membranes. While a large body of information exists on interactions of BAR proteins with membranes and regulatory proteins of the cytoskeleton, little is known about connections between these two processes. Here, we show that the F‐BAR domain protein pacsin2 is able to associate with actin filaments using the same concave surface employed to bind to membranes, while some other tested N‐BAR and F‐BAR proteins (endophilin, CIP4 and FCHO2) do not associate with actin. This finding reveals a new level of complexity in membrane remodeling processes.

Clustering Criteria and Algorithms

Cluster analysis is an unsupervised pattern recognition frequently used in biology, where large amounts of data must often be classified. Hierarchical agglomerative approaches, the most commonly used techniques in biology, are described in this chapter. Particular attention is put on techniques for validating the optimal cluster number and the clustering quality.

Proximity Measures for Cluster Analysis

The present chapter provides the basic information about the measures of proximity between two subjects or groups of subjects. It is obvious that these concepts must be clear in order to apply them to any pattern recognition analysis, both supervised and unsupervised.

Criteria to Extract High-Quality Protein Data Bank Subsets for Structure Users.

It is often necessary to build subsets of the Protein Data Bank to extract structural trends and average values. For this purpose it is mandatory that the subsets are non-redundant and of high quality. The first problem can be solved relatively easily at the sequence level or at the structural level. The second, on the contrary, needs special attention. It is not sufficient, in fact, to consider the crystallographic resolution and other feature must be taken into account: the absence of strings of residues from the electron density maps and from the files deposited in the Protein Data Bank; the B-factor values; the appropriate validation of the structural models; the quality of the electron density maps, which is not uniform; and the temperature of the diffraction experiments. More stringent criteria produce smaller subsets, which can be enlarged with more tolerant selection criteria. The incessant growth of the Protein Data Bank and especially of the number of high-resolution structures is allowing the use of more stringent selection criteria, with a consequent improvement of the quality of the subsets of the Protein Data Bank.

Silver and gold in the Protein Data Bank

The structural features of the silver and gold sites in protein crystal structures extracted from the Protein Data Bank have been investigated. It is observed that both cations have nearly always low oxidations states (+1) and low coordination numbers, adopt standard stereochemistries, and interact preferentially (particularly gold) with sulfur donor atoms of cysteine and methionine side-chains. Interestingly, gold cation have been very often refined with occupancy minor than 1.0 and are very often naked, in the sense that no donor atoms are sufficiently close to the metal cation. This apparently strange observation points out towards the need to develop specific and efficient validation tools for these elements when they are coordinated to proteins.

Protein hydration: Investigation of globular protein crystal structures

The positions of water molecules have been analyzed in high quality protein X-ray crystal structures. About 70% of these water molecules are in contact with protein atoms at the protein surface and constitute the first hydration layer. About 20% of them are close to the first hydration layer but are not in contact with protein atoms and constitute the second hydration layer. The rest of the water molecules are either buried in the protein core or close to hetero-atoms (inorganic ions and small organic molecules). Upper layers (third, fourth, etc.) are not observed in the dataset of protein crystal structures examined here. Water molecules of both layers are not, in general, surrounded by a tetrahedral arrangement of atoms, as it should be expected on the basis of the electronic structure of water. Usually there are less than four atoms around water molecules and even when there are four atoms, the stereochemistry is often distorted. Water molecules are more mobile than protein atoms, more in the second hydration layer than in the first.

Protomers of protein hetero-oligomers tend to resemble each other more than expected

A large fraction of the proteome is made by proteins that are not permanently monomeric but form oligomeric assemblies, which can be either homo- or hetero-oligomeric. Here it is described that protomers of hetero-oligomeric proteins tend to resemble each other more than expected. This is verified by comparing the level of similarity of pairs of hetero-oligomeric protein protomers and of pairs of proteins that do not interact with each other. This observation, interesting per se, might reflect the evolution of hetero-oligomers from ancestral homo-oligomers, through gene duplication and paralogs divergence. However, other hypotheses cannot be excluded and the observed structural similarity might result from several causes.

Structural features of uranium-protein complexes

Uranium toxicity depends on its chemical properties rather than on its radioactivity and involves its interaction with macromolecules. Here, a systematic survey of the structural features of the uranyl sites observed in protein crystal structures deposited in the Protein Data Bank is reported. Beside the two uranyl oxygens, which occupy the axial positions, uranium tends to be coordinated by five other oxygen atoms, which occupy the equatorial vertices of a pentagonal bipyramid. Even if one or more of these equatorial positions are sometime empty, they can be occupied only by oxygen atoms that belong to the carboxylate groups of Glu and Asp side-chains, usually acting as monodentate ligands, to water molecules, or to acetate anions. Although several uranium sites appear undefined or unrefined, with a single uranium atom that lacks the two uranyl oxygen atoms, this problem seems to become less frequent in recent years. However, it is clear that the crystallographic refinements of the uranyl sites are not always well restrained and a better parametrization of these restraints seems to be necessary.

How large B-factors can be in protein crystal structures

BackgroundProtein crystal structures are potentially over-interpreted since they are routinely refined without any restraint on the upper limit of atomic B-factors. Consequently, some of their atoms, undetected in the electron density maps, are allowed to reach extremely large B-factors, even above 100 square Angstroms, and their final positions are purely speculative and not based on any experimental evidence.ResultsA strategy to define B-factors upper limits is described here, based on the analysis of protein crystal structures deposited in the Protein Data Bank prior 2008, when the tendency to allow B-factor to arbitrary inflate was limited. This B-factor upper limit (B_max) is determined by extrapolating the relationship between crystal structure average B-factor and percentage of crystal volume occupied by solvent (pcVol) to pcVol =100%, when, ab absurdo, the crystal contains only liquid solvent, the structure of which is, by definition, undetectable in electron density maps.ConclusionsIt is thus possible to highlight structures with average B-factors larger than B_max, which should be considered with caution by the users of the information deposited in the Protein Data Bank, in order to avoid scientifically deleterious over-interpretations.