Beyond sequence similarity: Computational classification of plant enzymes through electrostatic similarity

Abstract

Here's the cleaned version:

Thesis significance statements

* Methodological Approach: This thesis applies a progression of computational classification methods to plant enzymes (phytoene synthase, DXO1, eIF4E, and the stress-sensing nucleotide phosphatases). Beginning with SSNs and Evolocity network analysis for large-scale classification, progressing through molecular phylogenetics for evolutionary inference, then structural similarity comparison, and culminating in a novel Fourier-decomposed electrostatic field comparison approach. Each method revealed limitations that motivated the next, with the electrostatic approach detecting functional relationships invisible to sequence and structural methods. * Key Evolutionary Discovery: Functional relationships can diverge from phylogenetic relationships. I show phylogenetically distinct proteins can share catalytic properties (Algal AHL and Ancient CNP clustering electrostatically with SAL), while phylogenetically similar proteins can be functionally divergent (TaSAL1/TaSAL2 with opposing yield effects; AHL-alpha/AHL-beta with differential pH sensitivity). * Biological Significance: Neo-functionalisation following whole-genome duplication is pervasive across plant protein families, with novel regulatory functions arising in DXO1, eIF(ISO)4E, and AHL sub-clades. The electrostatic clustering of Algal AHL and Ancient CNP with SAL suggests continuous PAP catabolic capacity across Viridiplantae, resolving an apparent functional gap spanning hundreds of millions of years where lineages lacking SAL appeared to have no enzyme for this essential function. This approach demonstrated that evolutionary classification can significantly inform decision making in breeding and genetic manipulation in crop species as demonstrated in SAL mutant wheat field trials.

Abstract

High-throughput sequencing has generated over 246 million protein sequences, yet fewer than 0.3% have experimental validation. This gap is acute in plants, where frequent whole-genome duplication creates paralogues that sequence-based methods cannot reliably classify as functionally equivalent or divergent.

This thesis applies a progression of computational classification methods to plant enzymes. Structural modelling of phytoene synthase variants elucidated mechanisms of enzyme dysfunction. Evolocity network analysis of DXO1 and eIF4E identified angiosperm-specific neo-functionalisation events. Sequence similarity networks combined with molecular phylogenetics classified the nucleotide phosphatase family into three distinct groups (SAL, AHL, CNP) and predicted functional divergence between wheat SAL paralogues, subsequently validated by field trials showing opposing yield effects.

Each method revealed limitations motivating the next, culminating in a novel approach comparing Fourier-decomposed active site electrostatic fields across 506 enzymes. Since catalysis depends on the electrostatic environment governing substrate binding, this captures functional information that sequence and structure cannot provide. Electrostatic analysis detected functional relationships invisible to other methods where Algal AHL and Ancient CNP clustered with SAL despite phylogenetic divergence, while AHL sub-clades showed regulatory divergence despite sequence conservation. Show more