Optical Microscopy in Chimera 2

Should Chimera 2 support analysis of optical microscopy data? Here are some of the considerations. April 30, 2014.

1. Chimera 1 has such minimal capabilities to work with optical microscopy data that it is seldom used for that purpose.

2. The main opportunity is to handle 3-dimensional optical microscopy. I think many programs to handle 2-d optical microscopy are well developed (e.g. ImageJ).

3. While Chimera 1 can read 3d optical microscopy images, the complexity of handling this data is that it is often a time series (imaging live samples) and often multi-channel. With the dimensions of time and fluorophore wavelength the data is often called 5-dimensional data. It requires user interfaces specifically designed for handling these extra dimensions.

4. Chimera 1 has the Volume Series tool for handling time-series of 3-dimensional images. It is a cumbersome user interface, not well-integrated with density map display. In 2013 I extended its capabilities to handle needs for Dyche Mullin's data of HL60 dendritic cells moving through collagen. Tom Ferrin showed this data at a talk at the NIH and had many inquiries afterward, for example, from Enrico Gratton who runs the Laboratory for Fluorescence Dynamics at UC Irvine.

5. Chimera 1 used to be able to display multi-channel image data in the Volume Viewer tool, developed for looking at fluorescently labeled chromosomes with John Sedat. That capability was removed around 2008 because it add a great deal of complexity to the user interface code and was seldom used. Multi-channel data can be displayed in current Chimera but with grayscale ("solid" style) rendering it does not blend the colors from the different channels. This is because Chimera in general does not correctly display 2 transparent models. While surface rendering works correctly for multichannel image data treating each channel as a separate map, surfaces often are not suitable for optical data because the intensity level varies such that there are interesting features at different levels that need to be shown simultaneously. In theory Chimera could blend the colors in grayscale mode for maps with identical grids without any change to user interfaces. This code would be moderately complex because it involves rendering multiple models as a single group contrary to the Chimera rendering paradigm where each model is rendered independently.

6. Optical microscopy, including the many super-resolution techniques, has too low resolution to be compared directly to molecular structures -- the core capability of Chimera. So adding optical microscopy capabilities would not benefit from the molecular analysis capabilities except to the extent that the same researchers also are interested in molecular structure and they could use the same software for both. The same criticism applies to serial milling electron microscopy techniques (focused ion beam SEM, and serial block face SEM), data we do try to handle since the electron microscopy Chimera user base also uses those EM techniques.

7. An enticing reason for developing optical microscopy analysis software is that 3d optical imaging is producing exciting data (e.g. brain imaging), the emerging microscope technology is rapidly improving and software to support it seems not to have been developed.

8. A key requirement to develop optical microscopy software will be collaborators with compelling data and a need for such software. At UCSF there are several researchers who might help. Professors John Sedat (now retired) and Bo Huang (Pharm Chem) develop super-resolution 3d microscopes. A fast (one 3d image per second) Bessel-beam microscope of the type used by Dyche Mullins at Janelia Farm is being built by Max Krummel (with Eric Betzig's guidance) for UCSF's Biological Imaging Development Center ​BIDC. Lastly, 3D confocal microscopy is quite common, but typically not as high resolution as Bessel Beam and has limited ability to capture dynamic data. Currently Dyche Mullins and Enrico Gratton are our active optical microscopy collaborators, but Max Krummel will likely be very interested as well once his microscope comes on line and the flood of associated data starts flowing. Dyche Mullins thinks that the microscope technology is no longer the bottleneck in science, but rather it's the visualization of the data. Here's a direct quote from him:

   “We are in the early days of working out the mechanisms and molecules that drive 3D cell migration
   and we are still limited by technology. We are less limited now by the technology of the microscopes
   than by the technologies of visualization. How do we turn the terabytes of data generated by a
   Bessel Beam into something that a human can understand and begin to theorize about?”
   

   The work being done in Dyche's lab makes for a great Driving Biological Project!

9. Measurement and segmentation would be the main analysis capabilities needed. Good visualization would need to be developed before those. To contrast Chimera's capabilites to other microscope software, take a look at Nico Stuurman's ​Micro-Manager project. Nico is a research associate in Ron Vale's lab. He had an R01 grant to support his work, but I believe the grant ended and was not renewed.

10. 5d optical microscopy data sets can be very large, for example Dyche Mullin's 3 minute cell motion data sets are about 50 Gbytes each (won't fit in memory), and often many data sets are acquired (I have about 1.5 Tbytes from Dyche). Handling this fits with our Chimera 2 emphasis on handling large data and also falls within NIH's recent "Big Data to Knowledge" (BD2K) program. Several RFAs have been released in 2014 under the BD2K umbrella, e.g., ​Development of Software and Analysis Methods....