Ab initio gene finding in maize. Training gene finders to annotate the maize genome.
Funded by NSF NSF-PGRP #0501758

 Project Objectives:

• Construct a "training-set" of maize gene sequence and use this to modify the TWINSCAN probability model. TWINSCAN is a dual-genome de novo prediction program that integrates traditional probability models with information from the alignments between two genomes.
• Benchmark TWINSCAN and compare its performance to that of other commonly used gene finders.
• Select a subset of maize hypothetical gene models predicted by TWINSCAN for wet-bench verification.
• Release the maize specific version of TWINSCAN to the scientific community as open source software.
• Use maize TWINSCAN to predict genes in available maize genomic sequence. Provide public access to the predictions.
• The bioinformatics analysis conducted during the course of this project is well suited as a training tool to future bioinformatics students. We have entered into an out-reach collaboration with Truman State University which includes training of student interns, seminars and invited lectures integrated into Truman's bioinformatics courses.

Experimental Approaches:

• Computationally identify high quality maize genes from public and proprietary collections of maize cDNA and genomic sequence. The identified collection of annotated maize genes will serve as a training set for modification of TWINSCAN.
• Use TWINSCAN to predict genes in available maize public genome sequence.a The results will be displayed in a web-based genome browser hosted at the Danforth Center.
• Benchmark TWINSCAN relative to other commonly used ab initio gene-finders, and re-train (when feasible) other open source gene prediction packages for maize.
• Wet-bench verify a subset of TWINSCAN predicted maize hypothetical gene models. Testing hypothetical predictions provides a second measure of TWINSCAN's specificity. Comparing experimentally verified models to predictions will effect further re-training.

Practical Applications of Research:

Maize genome sequence is the knowledge infrastructure for the next generation of plant molecular genetics and crop improvement. A broad understanding of the genes present in maize would provide the identities, and eventually the map positions, of many of the genes responsible for controlling agronomically important traits. Therefore, high throughput computational tools that can accurately identify genes within maize genomic sequence are absolutely necessary for annotating and understanding the maize genome. This project also illustrates the relative power of industry and academia collaborations focused on solving problems that plague research and development in both communities, and it serves as a paradigm for future industry-academia cooperative research activities.