Improved learning of Riemannian metrics for exploratory analysis.

We have earlier introduced a principle for learning metrics, which shows how metric-based methods can be made to focus on discriminative properties of data. The main applications are in supervising unsupervised learning to model interesting variation in data, instead of modeling all variation as plain unsupervised learning does. The metrics are derived by approximations to an information-geometric formulation. In this paper, we review the theory, introduce better approximations to the distances, and show how to apply them in two different kinds of unsupervised methods: prototype-based and pairwise distance-based. The two examples are self-organizing maps and multidimensional scaling (Sammon's mapping).     

A complementarity-determining region peptide of an anti-desmosome autoantibody may interact with the desmosomal plaque through molecular mimicry with a cytoplasmic desmoglein 1 sequence

The sera of patients with pemphigus, a group of autoimmune blistering skin diseases, contain autoantibodies directed against components of adhering junctions termed desmosomes. F12, a human monoclonal antibody derived from a pemphigus patient, recognizes an unknown polypeptide of the desmosomal and hemidesmosomal plaques. The third complementarity-determining region of the F12 heavy chain (VH-CDR3) was shown to share a four-amino-acid sequence (GSSG) with the intracellular domains of desmoglein 1 and bullous pemphigoid antigen 2 which interact with components of, respectively, the desmosomal and hemidesmosomal plaques. Computer modeling of F12 showed that the GSSG sequence protudes inside the antigen-combining site and thus might be involved in antigen interactions. The GSSG sequence is essential to F12 function, since a peptide containing the VH-CDR3 inhibited its binding to target antigens while VH-CDR3 peptides with specific modifications of the GSSG sequence did not. These data allow us to hypothesize that certain autoantibodies produced during the course of an autoimmune disease can behave as adhesion molecules, through the molecular mimicry of the motif involved in protein/protein adhesion, and to propose a new self-antigen binding mechanism for some autoantibodies.     

Quantification of in-plane motion of the coronary arteries during the cardiac cycle: Implications for acquisition window duration for MR flow quantification

Motion of the coronary arteries during the heart cycle can result in image blurring and inaccurate flow quantification by MR. This condition applies particularly for longer acquisition windows that are typical of breath-hold coronary flow measurements. To determine the sensitivity of the technique to in-plane motion of different coronary arteries, the temporal variation in coronary position was measured in a plane perpendicular to the proximal portion of the vessel. The results indicated the presence of substantial displacement of the coronary arteries within the cardiac cycle, with a magnitude of motion approximately twice as large for the right as for the left coronary arteries. An estimation of the resulting vessel blurring was calculated, showing that the duration of the acquisition window for high spatial resolution coronary flow acquisitions should be less than 25 to 120 msec, depending on the specific coronary artery studied. In addition, these data specify optimal acquisition window placement for high resolution coronary angiography.