v2.9.3

• Significant additions have been made to the function (slml2report.m) for comparing two CellOrganizer spharm-obj models of organelle shape and spatial distribution. The report now includes:
  -- improved plots of the spatial distributions of objects, and
  --a statistical comparison between the positions of objects (relative to cell and nuclear boundaries) in the two models. This is done by forming a graph of the positions of all objects in both models, finding cliques within that graph, and measuring the fraction of objects from each model in each clique. This is used to calculate the KL divergences between two random halves of the objects in each model (to establish a baseline measure for the single models), and between the four cross-model combinations of halves. The higher the divergences of the cross-model pairs are compared to the within-model pairs, the more different the spatial distributions of the models are considered to be.
• The execution time of the SPHARM-RPDM code for constructing cell, nucleus and organelle shape models has been improved through more efficient code for parts of the shape parameterization step contributed by Khaled Khairy (khaled.khairy@stjude.org). Khaled is the Director of the Center for Bioimage Informatics at St. Jude Children’s Research Hospital in Memphis. We gratefully acknowledge his contribution. (EqualAreaParametricMeshNewtonMethod.m, calc_area_jacobian_sphere_KK.m)
• Improve object detection for spharm_obj models. (find_objects_local.m, find_objects_local.m, spharm_obj_model.m, spharm_obj_percell_3D.m)

v2.9.2

Bug Fixes

• Fixes to functions img2slml in param2model_3D in which the correct pathway for spharm obj models weren’t properly handled.
• Other minor pathway and integrity fixes to slml2info.

v2.9.1

New features

• Added new 3D model type (class “vesicle” and type “spharm_obj”) for organelles; it models organelle size and shape as well as subcellular position. “Vesicle” is the CellOrganizer class that includes all organelles that consist of discrete objects. The previous models in the “vesicle” class used ellipsoids to model individual objects. The new model is learned by first segmenting all objects in a protein image using adaptive thresholding, and the position of each object relative to the cell and nuclear membranes is recorded. A size and shape model is then constructed from all objects using the SPHARM-RPDM method, and a logistic regression model is learned to capture the probability density of an object occurring at any position in a standardized cell. Synthesis from the learned models is not yet supported, but strong support is provided (through slml2report) for comparing models between datasets. New demos demo3D61 and demo3D62 illustrate training this new model type.
• Added new functionality to slml2report to provide comparison of SPHARM models.
• Added support for exporting synthetic cell instances as VCML files for use in Virtual Cell.

Enhancements

• Support for exporting instances as SBML Spatial files has been updated to the latest version of the standard. Demo demo3D64 illustrates the creation of SBML spatial files.