POGIGWASC

A gene predictor developed on Loxodes magnus in the summer-semester of 2020 as part of the master-thesis of David Emanuel Vetter (Thesis-title: "Prediction of genes in genomes with ambiguous genetic codes"). Heavily based on Augustus.

Naming: POGIGWASC (IPA: [ˈpɔ.dʒɪdʒ.wəskʰ]) stands for "Prediction of genes in genomes with ambiguous stop codons" -- the deviation from the thesis-name is due to the program only being specifically designed to handle an ambiguous UGA-codon. Its generalizeability to other cases of genetic code ambiguity is unknown.

Running the program:

Variant 1: Without eclipse -- just want to use the program

Prerequisites: install Maven

1. download the code as a zip-file (from github) and extract the contents to some directory
2. navigate to that directory (more precisely: navigate to the directory masterthesis-master, containing the pom.xml) and run the following maven command: mvn package appassembler:assemble (this should print quite a bunch of information into the console; including (green) text informing about the number of tests run, and the number of failures encountered -- there should not be any Failures or errors; at the end, 'BUILD SUCCESS' should be printed to the console)
3. from the same directory, run target/appassembler/bin/ghmm-predict: this should print the standard help-text

To move the compiled program, copy/move the entire appassembler-directory (which can be renamed) -- it does not suffice to copy/move just the generated target/appassembler/bin/ghmm-predict file

Variant 2: For working on the program

Prerequisites: install Eclipse

1. clone the git-repository with eclipse (or directly with git)
2. right-click the pom-xml and choose Run As > 2 Maven build..., then enter e.g. package appassembler:assemble for the goals and run This should produce an output as described above
3. Rightclick de.vetter.pogigwasc/App.java and run as java application. This should give the usual help-text (set command-line args via Run>Run Configurations)

Program structure overview:

Directly in src/main/java/de/vetter/pogigwasc, most relevant classes for gene-prediction are found, in particular, the main-class App with the main-method that defines what happens, when the program is run (cf. pom.xml).

• ModelParameters is used to load and query the model-parameters from an external file
• Viterbi and ViterbiSeed implement the viterbi-algorithm
• GHMM implements GHMMs
• LoxodesMagnusGHMM holds the exact model used for Loxodes magnus (and extends GHMM)
• LoxodesMagnusIntronless holds the model used for intronless predictions in Loxodes magnus
• Pair and GFFFeature are very small and trivial (implementing a pair, and being an enum for writing GFF3-files)
• The folder/subpackage states/ contains the implementations of the states: These take care of emission probabilities and enumerating valid emission lengths

The Classes are documented in more detail in their respective files. Tests are found in src/test/java/de/vetter/pogigwasc, where running AllTests runs all tests.