Logan Johnson

 

Systems, Image Analysis, Web and Database programming/development, Server administration, Grid computing

I am the liaison between the Spalding lab's Phytomorph project and the Condor team headed by Miron Livny, Professor of Computer Sciences. I am responsible for getting custom code to run correctly on machines in the UW Center for High Throughput Computing. I manage our servers and storage arrays and provide IT support to our lab. I also coordinate the data flow between our collaborators and us.

I also develop image analysis algorithms in matlab for automating measurements on time series imaged roots to identify phenotypes in the early development stages of Arabidopsis. In the past I have worked on algorithms to measure root length and root tip angle of roots responding to gravitropism. Currently I am in development on an algorithm to extract measurements on branching root structures, such as growth rate, length, tip angle, and distance down primary root, among others, from lateral roots. I developed this algorithm for Arabidopsis roots, but could potentially be ported to work with other species. I also support, update and fix a number of other image analysis algorithms developed by the Spalding lab.

I also am developing a web application and database so botanists can more easily track genetic lineage, as paper and pencil becomes antiquated with high throughput image acquisition and extraordinarily large datasets. This application will also help scientists seamlessly combine metadata and measurements with visualization tools to better detect patterns. This web app will be extended to seamlessly continue statistical analysis of the measurements, either through a simple feature rich export process or use of standard statistical methods on the data within the web app, this could include more complex statistics such as quantitative trait loci (QTL) analysis. Though the web app may focus on seedlings as that is where most of the significant measurements to us can be taken, we will continue to embed the three major phases of plants, seed, seedling and mature plant, collating all the meta data and measurements we can about each. It will eventually become a complex tree of genealogy, with the mature plant associated with a seedling and a seedling associated with a seed and a seed associated with a mature plant. The process will be double redundant as each sample will have a unique id, correlating to a database entry with a virtual genealogical tree using database links. In addition to this there will be a micro barcode associated with each sample, that contains (in encoded text) one generation of meta-data and a link (ID) to the previous generation. Thus one could reconstruct the history even with a computer failure. This is important as previously with the paper and pencil method, you had a physical copy of the information for scientists security and sanity. There is a significant advantage to loading the database with meta data, it makes complex queries effortless, collating of data that would take weeks, will now take a matter of minutes. Part of the goal of the micro barcoding is to create a high capacity, redundant, low cost, barcoding system high throughput plant experiments. All of this started as a way to process the output of our high throughput phenotyping robot, developed by Ram Subramanian.

 

Spalding Lab

Logan Johnson

I work on the Phytomorph project.

CHTC

The Center for High Throughput Computing was established in 2006 to meet the advanced computing needs of researchers.

Condor

What is Condor?