Projects
The breadth of our research
Our research covers key areas in AI-based discrete optimisation.
MiniZinc
Project Lead: Dr Guido Tack
MiniZinc is a free and open-source constraint modelling language developed in collaboration with Data61 Decision Sciences and the University of Melbourne.
It can be used to model constraint satisfaction and optimisation problems in a high-level, solver-independent way, taking advantage of a large library of predefined constraints. The model is then compiled into FlatZinc, a solver input language that is understood by a wide range of solvers.
This project continues to develop and extend the MiniZinc language, making it easier to use and more efficient with greater applicability to new problem classes. Current research spans areas such as Stochastic Optimisation, Large Neighbourhood Search and debugging and profiling constraint models.
Learning from learning solvers
Project lead: Professor Maria Garcia de la Banda
Combinatorial optimisation problems are solved by finding a combination of choices that satisfies a set of constraints and optimises an objective function. Unfortunately, solutions are remarkably difficult to obtain due to the exponential number of possible choice combinations.
The goal of this project is to revolutionise the modelling of combinatorial optimisation problems thanks to a new insight: powerful technology developed to solve (rather than model) these problems can also be used to improve the model itself.
Next-generation protein structural alignment using information theory
Project lead: Dr Arun Konagurthu
Progress in protein structural biology has radically transformed our understanding of cellular and biological processes, driving far-reaching research advances in life sciences and medicine. Importantly, these advances rely heavily on a small set of computational technologies, of which structural alignments play an indispensable role.
A structural alignment is the assignment of correspondences between the 3D coordinates of the amino acids of two or more proteins. Among many diverse uses, structural alignments are crucial in unravelling the distant past of protein evolution, unreachable by any other means
There are many methods to produce structural alignments for a given set of proteins. But several studies have shown that the alignments obtained by these methods differ substantially.
This project aims to address the deficiencies in the current structural alignment state of the art by using:
- information theory to develop a framework of measuring structural alignment quality
- modern combinatorial optimisation technology to design new search algorithms capable of obtaining biologically-meaningful and optimal structural alignments
- visualisation technology to provide innovative ways of exploring, visually and quantitatively, the quality of competing alignments for a given set of structures.
Understanding the architectural principles of protein structure
Project lead: Dr Arun Konagurthu
We are investigating how to unravel the observed repertoire of protein folding patterns and working on identifying a universal set of architectural themes or concepts – essential to understanding how protein 3D shapes form, function and evolve.
Using statistical learning, information theory, and optimisation our recent work identified a comprehensive dictionary of concepts into which any protein folding patterns can be decomposed.
By decomposing the entire world-wide Protein Data Bank using the dictionary, we are able to understand the constituents of protein folding patterns, establish their functional roles and reveal patterns of conservation of amino-acid sequences that dictate protein 3D shapes.
The insights gained from this investigation are useful for annotation of protein function, protein engineering, drug design and protein structure prediction.
Statistical inference of protein structural alignments
Project lead: Dr Arun Konagurthu
Comparing the 3D structures of proteins and establishing their relationship is a computational task that supports many biological studies.
In this ARC-funded project we are investigating statistically rigorous ways to assess and identify biologically meaningful structural relationships between proteins.
We recently developed an information-theoretic measure to accurately discriminate between competing structural alignments, and select the best by establishing an objective trade-off between alignment complexity and structural fidelity.
Among other outputs, this research has resulted in the program MMLigner that’s unlike existing alignment programs. It’s able to infer closely-competing and statistically-significant structural alignments between proteins, instead of reporting just the single best.
Computational methods to study of protein structure and architecture
Project lead: Dr Arun Konagurthu
We have contributed to the theory and practice of algorithms that can compare and classify protein structures.
The algorithms are fundamental to understanding the evolution of protein families and identifying patterns of conservation in sequences that fold into similar structures.
Some of our popular work that is used by structural biologists and crystallographers around the world are: