next up previous contents
Next: Current State of Protein Up: Introduction Previous: Goals

Motivation

The motivation for this work falls into at least two categories: numerical description of protein localization and classification of protein localization.

Currently, assessment of subcellular protein localization is subjectively carried out on an investigator by investigator basis. By describing the localization of proteins numerically, we will have another objective way of describing proteins in addition to those methods that exist now (e.g., amino acid sequence, hydrophobicity, functional motifs, etc.) Furthermore, we will be able to introduce some degree of formality to the process of describing the localization of a protein. Ultimately, it will be possible to quantitate the degree of similarity in the localization of two proteins, just as it is now possible to quantitatively describe the degree of similarity in amino acid sequence. A benefit of such quantitative analysis will be the ability to obtain novel information about new or existing proteins; a list of proteins with the same or similar localization characteristics, for instance. Localization information will be archived in a database so that it can be retrieved as easily as DNA and protein sequences are today.

The ability to classify protein localization patterns is a direct benefit of being able to numerically describe them. By generating numbers that are sufficient to describe a pattern, we are simultaneously generating features that can be used in any of a variety of pattern recognition schemes. Some applications that will see immediate benefit from this work include high throughput screening, microscope automation, and automated gene discovery based on protein localization.

Automated screening of microscope images is becoming an increasingly important tool in a variety of fields including biology and pharmacology. Pharmaceutical companies have a large arsenal of compounds, perhaps hundreds of thousands, that are potentially marketable drugs. It is a significant effort to identify those few that have a desired effect on a system (e.g., those compounds that prevent translocation of a transcription factor to the nucleus). Automated screening of protein localization patterns is a potentially very useful adjunct to this process.

A closely related application is microscope automation. Instead of manually scanning an entire sample for those cells that are of interest, it will be possible to have the microscope hardware find those cells and then image or interact with them at an appropriate time. Technology such as this will be particularly useful when studying events that constitute a minority of the cell cycle, phenotypes present in only a few cells, or phenomena that are otherwise rare.

Finally, there is the potential to help automate gene discovery based on the localization of the corresponding protein. Investigators interested in a particular subcellular structure (i.e., organelle or compartment) will be able to identify heretofore undiscovered genes that code for proteins that localize to that structure. There is, of course, considerable molecular biology involved in such a system and no absolute need for automated pattern recognition, but automated screening for patterns of interest will speed the process many fold. Such an increase in throughput will be essential as the known complexity of the human genome grows.


next up previous contents
Next: Current State of Protein Up: Introduction Previous: Goals
Copyright ©1999 Michael V. Boland
1999-09-18