Prediction Informatics Resources And Techniques

This page is a repository for gene finders and other bioinformatics prediction softwares developped by us and our collaborators. Included are full source code for all our gene-finders as well as C++ class libraries and other components for prediction and machine learning tasks.

See also http://www.

Source Code -- Complete Projects

Project

Description

Language

TigrScan

GHMM gene-finder like Genscan/Genie written in highly optimized C++ & designed to be extensible and reusable for other tasks related to gene-finding. Uses IMMs.

C++

GlimmerHMM

GHMM gene-finder like Genscan/Genie, which makes use of the techniques implemented previously by GlimmerM : OC1 decision trees and IMMs. Fast and accurate.

JIGSAW

Uses the output from gene finders, splice site prediction programs and sequence alignments to predict gene models.

GlimmerM

Eurokaryotic gene-finder using OC1 decision trees and Interpolated Markov Models.

Unveil

Pure HMM-based gene-finder based on the VEIL model. Highly optimized C++.

C++

ELPH

Gibbs sampler for finding motifs in DNA; has been used for detecting exon splice enhancers (ESE's). Also applicable to other motif-detection tasks.

GeneSplicer

A fast, flexible system for detecting splice sites in eukaryotic DNA.

Glimmer

Prokaryotic gene-finder using IMM's.

TransTerm

A program that finds rho-independent transcription terminators in bacterial genomes

RBSfinder

A program to find ribosome binding sites in prokaryotic DNA.

Perl

VEIL

The original VEIL gene-finder.

C++

MORGAN

A eukaryotic gene-finder using OC1 decision trees.

All of the software listed above is Open Source and is distributed under the ARTISTIC LICENSE. See www.opensource.org.

Source Code -- Reusable Software Components

Package

Description

Language

OC1

oblique decision trees for classification

NET

backpropagation neural networks for classification

C++

entropy-based decision trees for classification

C++

Suffix Trees

Suffix trees by Stefan Kurtz

tigr++

C++ class library used by several TIGR genefinders and other packages. Covers string & sequence processing, math/statistics, many efficient data structures, GFF parsing, sorting, and I/O.

C++

regress

multivariate regression for classification

C++

bayes

Naive Bayes classifier

C++

genetic algorithms / genetic programming

C++

KNN

K-nearest-neighbors classifier with Mahalanobis distance (to control for correlation among attributes) and feature selection based on F-ratio.

C++

All of the software listed above is Open Source and is distributed under the ARTISTIC LICENSE. See www.opensource.org.

Documentation
TigrScan User Manual	How to install and use the TigrScan gene-finder
TigrScan Training Manual	How to train the TigrScan gene-finder
TigrScan Software Architecture	How the TigrScan gene-finder software is structured -- for those who wish to modify the program
TIGR++ Reference Manual	Class reference
Machine-learning packages	Brief description of the bayesian / neural / tree / regression / nearest-neighbor / genetic-algorithm packages listed above

Training Data
Arabidopsis.thaliana.tar.gz	GFF coordinates & FASTA file
Aspergillus.fumigatus.tar.gz	GFF coordinates & FASTA file
Aspergillus.spp.tar.gz	GFF coordinates & FASTA file
Homo.sapiens.tar.gz	GFF coordinates & FASTA file
Mus.musculus.tar.gz	GFF coordinates & FASTA file
Plasmodium.falciparum.tar.gz	GFF coordinates & FASTA file
sample-prediction-problems.tar.gz	sample training/test/configuration files for machine-learning packages

Experimental Results
description	poster	date
A comparison of machine learning methods for discriminating exons from non-exon ORFs	Bill Majoros	1-19-04
Some accuracy/speed/memory-usage statistics for TigrScan	Bill Majoros	2-10-04
Results of running a Plasmodium-trained genefinder on Tetrahymena	Bill Majoros	2-13-04
Interpolated vs. Non-Interpolated Markov Chains	Bill Majoros	3-3-04

See also http://www.