Recently, employing multivariate statistical techniques for data, that are geometrically random, made more attention by the researchers from applied disciplines. Shape statistics, as a new branch of stochastic geometry, constitute batch of such data. However, due to non-Euclidean feature of such data, adopting usual tools from the multivariate statistics to proper statistical analysis of them is not somewhat clear. How to cluster the shape data is studied in this paper and then its performance is compared with the traditional view of multivariate statistics to this subject via applying these methods to analysis the distal femur.

In this paper, we study the applicability of probabilistic solutions for the alignment of tertiary structure of proteins and discuss its difference with the deterministic algorithms. For this purpose, we introduce two Bayesian models and address a solution to add amino acid sequence and type (primary structure) to protein alignment. Furthermore, we will study the parameter estimation with Markov Chain Monte Carlo sampling from the posterior distribution. Finally, in order to see the effectiveness of these methods in the protein alignment, we have compared the parameter estimations in a real data set.