SegaPred: Protein Structure Prediction Software
I developed a protein structure predicting program named SegaPred with Prof. Jie-Oh Lee of KAIST during September 2001–February 2002. Using Self-Evolutionary Genetic Algorithm, CHARMM potential, and some of the NAMD code, we were able to demonstrate the software can predict some model cases quite accurately (Fig. 1).
Figure 1: superimposed native (blue) and predicted (red) peptides. (a) A predicted 12-mer poly-alanine molecule with 0.327-angstrom all-atom RMSD. (b) A predicted crambin molecule with 2.695-angstrom all-atom RMSD.
Some notable things
Contrary to many other protein folding programs, SegaPred does not rely on molecular dynamics simulations. Instead, it predicts the native structure of proteins using Self-Evolutionary Genetic Algorithm (as its name implies). I designed and implemented the most part of SegaPred. In particular, I implemented the genetic algorithm module and its mutation operators (Fig. 2).
Figure 2: mutation operators used in SegaPred. (a) The original conformation. (b) Rotation of dihedral angles: the 
of Ile rotated. (c) Rotation of C’ bond: the C’ of Ile rotated. (d) Change of displacement: the length of the 
bond changed.
Manuscript
I was awarded Samsung Humantech Thesis Prize (Silver Prize of the High School Division) for this work, and it was published as: Ha Hong, “SegaPred, A Self-Evolutionary Genetic Algorithm Based Protein Conformation Predicting Program,” Samsung Humantech Thesis Prize 8 (2002). This became my high school thesis.
Follow-up researches
- Eigenmode Analysis of Enzyme Carbon Backbones around the Active Sites
- Parallel Programming with MPI and its Application to SegaPred
