Biomedical subjects

Pier Luigi Bellon

Publications and source records attributed to Pier Luigi Bellon.

4 recordsLinked to original sources

PubMed2004

VIVA (from virus variance), a library to reconstruct icosahedral viruses based on the variance of structural models.

VIVA is a software library that obtains low-resolution models of icosahedral viruses from projections observed at the electron microscope. VIVA works in a fully automatic way without any initial model. This feature eliminates the possibility of bias that could originate from the alignment of the projections to an external preliminary model. VIVA determines the viewing direction of the virus images by computation of sets of single particle reconstruction (SPR) followed by a variance analysis and classification of the 3D models. All structures are reduced in size to speed up computation. This limits the resolution of a VIVA reconstruction. The models obtained can be subsequently refined at best with use of standard libraries. Up today, VIVA has successfully solved the structure of all viruses tested, some of which being considered refractory particles. The VIVA library is written in 'C' language and is devised to run on widespread Linux computers.

Analysis of Variance↗

PubMed2003

The variance of icosahedral virus models is a key indicator in the structure determination: a model-free reconstruction of viruses, suitable for refractory particles.

A model-free method to determine the three-dimensional structure of icosahedral viruses is described. The novel strategy is based upon the approximate principle that correct virus structures have high variance as do all other well-detailed structures, even wrong ones. The original projections of individual particles are reduced to a radius of 25 pixels and are used to compute single particle reconstruction models by assigning them 1800 different Euler triads. The variance of the models obtained from all projections is stored in maps and a decimation process is carried out. In a first stage, thresholds are adopted for the variance values, and in a second stage, carried out by correspondence analysis and classification, 30 clusters of models are sorted out. The clusters are refined to yield models contained in boxes of 64(3) voxels. The refined models with highest variance and closest similarity represent the correct solution. Once enlarged, these models can be used to align all available projections in their original scale in a customary projection-matching process. The method has proved successful in determining the structures of poliovirus, of the empty and filled capsids of L-A virus, and of a modified capsid of hepatitis B virus.

Capsid↗

PubMed2002

A trajectory-based algorithm to determine and refine Euler angles of projections in three-dimensional microscopy. Improvements and tests.

An improvement of the trajectory matching algorithm is presented, which is based on the use of the derivative of trajectories and of the projection of experimental sinogram lines in the factor space determined by sinogram lines of projections of a model. The algorithm performance is illustrated by use of different phantom structures, to show the effect of symmetry on trajectory matching. A GroEL complex has also been reconstructed from both cryo-negatively stained and unstained frozen-hydrated samples. The refinement of this structure has been carried out by the trajectory matching algorithm as well as by conventional cross-correlation methods. Slight differences among the two results are discussed. The improved trajectory matching algorithm, based on chi2 distances, runs much faster than correlation analysis and looks satisfactory as for the quality of the reconstructed structures.

Journal Article↗

PubMed2002

Symmetry embedding in the reconstruction of macromolecular assemblies via the discrete Radon transform.

In this paper we discuss the embedding of symmetry information in an algorithm for three-dimensional reconstruction, which is based on the discrete Radon transform. The original algorithm was designed for randomly oriented and in principal asymmetric particles. The expanded version presented here covers all symmetry point groups which can be exhibited by macromolecular protein assemblies. The orientations of all symmetry equivalent projections, based on the orientation of an experimental projection, are obtained using global group operators. Further, an improved interpolation scheme for the recovery of the three-dimensional discrete Radon transform has been designed for greater computational efficiency. The algorithm has been tested on phantom structures as well as on real data, a virus structure possessing icosahedral symmetry.

Algorithms↗