Adaptively-Branching Fuzzy Greedy K-mean Decision Forest (FGK-DF) Model for Protein Local Tertiary Structure Prediction

For the past twenty years, protein tertiary structure research has been given much attention, with solutions existing for both wet lab procedures (x-ray crystallography and NMR spectroscopy) and bioinformatics approaches (threading, homology-modeling, and de novo). Unfortunately, each approach has significant shortcomings, such as necessary time, capital, expertise (for wet lab procedures) or restrictions imposed by the method, limiting the resolution or novelty of produced tertiary structures (for bioinformatics approaches). This work propose the Adaptively-Branching Fuzzy Greedy K-means-Decision Forest (FGK-DF) model, which utilizes conserved sequential and structural motifs that transcend protein family boundaries, to predict the local tertiary structure of proteins with unknown structures. In this work, the FGK-DF model is conceptually compared against existing approaches and explicitly compared against the Super Granule Support Vector Machine approach (Super GSVM), with accuracy and coverage results highlighted.