Intelligent Data Collection and Analysis
Our move to high-content screening with the Panoptes instrument requires shifting away from our philosophy of providing the scientist with real-time feedback of all available data during an experiment – doing so would increase by 100x the time taken for an experiment. In the first step, the data has to be collected, analyzed, and compressed without scientist interaction. Subsequently, the data needs to be summarized. Some of our recent and ongoing work is described below.
Experiment Specification: The first step in an experiment is providing specimen information (tissue type, added enzyme concentrations, etc.) and experiment parameters (number of locations to search, how long to dwell at each, etc.) The figure below shows our new Well-Layout tool. Together with an experiment configuration file, this tool fully specifies the data collection and drives the analysis.
Experiment Execution and Analysis: Our current command-line application called Panoptic Nerve runs on the cluster of Panoptes’ control computers to coordinate data collection on the system. It implements all phases of the passive bead experiment workflow used in most Panoptes experiments, including:
- a Z/tilt calibration routine to locate the bottom of the plate and bring it into range for all 12 cameras,
- autofocus routines that locate the best-focus plane within each field of view,
- optimized bead detection and tracking routines to provide 2D tracks of beads over time,
- Matlab-based Mean Squared Displacement (MSD) calculations to determine viscoelasticity.
We are moving to a new graphical experiment-description tool named PanoptesRun that handle the creation of output directories and generation of run-configuration files. It also optimizes the layout of regions to be studied within each well. The figure below shows a prototype of the tool.
LEWOS: To enable all twelve wells to work in coordination, and to enable maximum efficiency in data collection, we have developed a Launching Events With Optimal Synchrony (LEWOS) architecture that provides synchronized control of cameras, LEDs, and focus on the multiple channels. We also developed custom microcontroller hardware to implement this architecture.
We have used (and continue to use) a number of commercial and open-source instrument-control architectures in our lab including Metamorph, IPlab, Matlab, and Micro-Manager, and contributed updates to Micro-Manager to enable control of our specific devices. These work well for control of multiple devices attached to a single computer and for synchronizations of devices on the order of fractions of a second to seconds (in fact, the operation of motor-driven devices over RS-232 often takes this long for operations to complete). We have also developed multi-computer barrier-synchronized data collection based on our VRPN technology. This enabled us to synchronize devices between computers to the same degree.
The Panoptes system requires sub-millisecond synchronization among five or more channels (camera, three LEDs, electronic focus, magnetic or electrical signals to apply to wells) on twelve channels, including the ability to repeatedly run complex sequences of illumination, image acquisition, and focus control at full video rates. The control sequences must be flexible enough to change based on image information and dynamic changes in experiment protocol. This required us to develop the LEWOS architecture to drive the custom array of microcontrollers on Panoptes.
LEWOS provides microsecond specification of camera exposure start/stop, LED turn on/off, varioptic focus setting, and other parameters. It includes a global timing domain a well as twelve local timing domains to enable both within-channel and between-channel synchronization depending on the needs of a particular experiment. To enable both flexibility and accurate timing on sequences of operations, it includes a macro-specification capability so that specific protocols needed for a given experiment can be loaded and then triggered. Significantly, every instruction in LEWOS includes a time-stamp, which can be with respect to global time, local time, or the time of the previous instruction. This enables both flexibility and accurate timing, providing an effective real-time operating system on which to run experiments.
LEWOS is available for licensing; contact us if you are interested.
The data acquired from Panoptes is compressed by a scientific loss-less compression algorithm developed in our lab, which is published in 2014. The compression ratio of the algorithm is up to 100:1. Meanwhile, we are investigating a new method that will achieve an even higher compression ratio than the current one. Also this new compression should not greatly impact the video quality in terms of scientific analysis.