Computational Biology Project

Summary

Biological research today is facilitated by automation that provides instrumentation control and data acquisition. Researchers are now able to quickly collect large amounts of image-based data that serves as the primary output of their experiments and as the source of their measurements. The biological researcher is left with huge amounts of image data to process and analyze using techniques that are usually outside of their field of expertise. In addition, the large amounts of raw and processed data require large amounts of metadata for their correct interpretation, handling and storage.

The literature is replete with references to and descriptions of image processing techniques, but experience shows that many techniques have limited applicability. Certain techniques work well only on certain types of images; two images from different data channels of a microscope may require fundamentally different techniques. Typically, researchers choose tools and techniques that they have been exposed to and feel comfortable with. There is little guidance available and much of the biological literature seems to give little information about the methods used for analyzing experimental data and their associated parameters.

A basic tenet of this project is that image processing and analysis techniques, despite their implementation in software are fundamentally measurements and not simply calculations. This suggests that the measurement uncertainty associated with the use of software-based image processing and analysis methods can and should be determined. It also suggests that clear guidance can be given to users as to the applicability of various techniques.

The increasing prominence of biological image-based data and the importance of biological metadata raise questions related to information technology standards. The biological research community is currently beset with a plethora of disparate image and data formats and the same can be expected once metadata management solutions become increasingly available. The ability to perform inter-laboratory comparisons, a necessity for ensuring measurement consistency, is also hampered and needs to be addressed.

Description

Overview:

• High-throughput technologies are generating large amounts of complex data that is difficult to process and convert into knowledge
• Issues exist throughout data lifecycle:
• • Acquisition
  • Analysis
  • Archiving
  • Interchange
• Imaging technologies are increasingly being used both as diagnostic tools and as research tools in the biosciences
• • Novel methods are needed for automated analysis and comparisons
  • Correlation among studies is difficult at best
  • Off-the-shelf methods are not well characterized and can contribute significantly measurement uncertainty
  • Need to combine features from images with other biological or medical sources

NIST Role:

Provide the measurement tools and standards that enable quantifiable and reproducible measurements of cells and their interactions through:

• Standard data/metadata format for image capture, storage, retrieval, analysis
• Software to enable high throughput cell image analysis and interoperability
• Standards and validation required to ensure reproducible image analysis
• Multi-site experiments to test software and validation protocols

Technical Approach

• Create and evaluate an integrated data collection, organization and analysis infrastructure for cellular imaging
• Experimentalists and computational scientists - focus on the physical standards and protocols for data collection, image processing and analysis, storage of data and metadata, and evaluation of results