Technical Highlight - February 2015
Short description: The cellPACK software generates three-dimensional mesoscale models for a detailed look at biological systems like the cell.
Methods in structural biology allow us to zoom in on molecular details, and microscopy gives a view of the cell, but there is often a gap in our conception of how the two scales fit together. To bridge the dimensions of molecular and cellular biology, Johnson, Olson and colleagues (PSI CESG) have developed the flexible probabilistic modeling software cellPACK.
The program builds on their open-source autoPACK, a more general approach to the problem of loose packing irregular objects within or around a defined volume or surface. AutoPACK finds a heuristic solution to packing objects of arbitrary shape with minimal overlaps, producing a distribution with user-defined interactions and constraints. It also accommodates procedural components such as fibrous or branching growth that must be assembled in a specific order.
Biological systems like the cell are not empty sacks that need filling. Ultrastructural elements such as membranes and fibers provide additional levels of structure. CellPACK is a biology-specific extension of autoPACK that contains pre-loaded 'recipes', such as synaptic vessels or virions, that include surface meshes for structural components such as organellar outlines derived from tomographic data, to place tens to millions of 'ingredients', i.e. molecular models with appropriate localizations.
These tools automate pipelines for ingredient extraction and packing, greatly reducing the amount of hands-on time needed to generate models. They also allow editing in a user-friendly graphical interface or as Python script: recipe components, molecular densities and behaviors, visualization and animation are all highly customizable and accessible to both novice and advanced users. A suite of analysis tools helps to validate distributions and detect unexpected biases.
CellPACK works at the mesoscale, approximately 10–100 nm. The authors produced a set of models of increasing complexity: unstructured blood plasma, a 100-nm cube of synaptic vesicles, a model of HIV in blood plasma and a whole-cell model of Mycoplasma mycoides. The modeling of HIV-1 helped to distinguish between competing models of Env surface protein distribution based on conflicting experimental data. An important feature of the open-source tools is the ability of experts in the community to continually update and improve recipes.
By synthesizing information collected at different scales, cellPACK and autoPACK let us explore a more holistic picture of biology.
G.T. Johnson et al. cellPACK: a virtual mesoscope to model and visualize structural systems biology.
Nat. Methods 12, 85-91 (2014). doi:10.1038/nmeth.3204