Winkler – Probing materials-biology interactions using AI and machine learning
Computation is the third arm of research, after theory and experiment. Computation and simulation of molecular systems is becoming indispensable for 21st century science. However, the size, scale and complexity of realistic materials-biology interactions precludes the application of rigorous, physics-based computational methods like molecular dynamics and quantum chemistry. AI and machine learning are making spectacular inroads into solving some of these very complex problems.
We use a wide range of computational chemistry and AI-based methods (principally machine learning) to model complex systems to predict their properties and gain insight into mechanism of interaction at the molecular level. As these are broadly applicable platform methods, we collaborate with experimental scientists across a wide range of projects, some of which address non-biological structure-property relationships problems in materials.
Research areas
Machine learning for materials and surface science
We have applied advanced informatics and machine learning methods to extract new knowledge from surface analysis methods. We are applying these methods to tissue profiling or tumour samples (Gardner et al. Anal Chem 2020) and to libraries of biomaterials that are candidates for coatings for implantable and indwelling medical devices (Burroughs et al. PNAS, 2020).
Next Generation Biomaterials
Working with the University of Nottingham on a large EPSRC project discovering and designing new materials for medical applications. We use data from high throughput experimentation to build predictive models of the biological effects of biomaterials. We have capabilities to explore a large range of surface chemistries and microtopographies and to capture them as novel mathematical descriptors for training machine learning models (Vallieres et al. Science Adv. 2020; Celiz et al. Nature Mater. 2014).
Therapeutic gases
Noble gases are chemically inert but display a wide range of potentially useful biological effects. Working with the French multinational, Air Liquide Santé International, we use large scale computational simulations to understand the molecular bases for these medically relevant properties (Winkler et al. J. Chem. Inf. Model. 2019). We intend to develop delivery systems for noble gases that will improve their efficacy and make selective targeting possible.
2D and porous materials
We use machine learning and evolutionary methods to design and optimise porous materials for environment and energy applications (Thornton et al. Chem. Mater. 2017), and to model the properties of hybrid 2D materials for electrooptical applications. We generate predictive models of interlayer spacings in hybrid 2D materials and predicted band gaps and super lubrication effects of these novel materials, recently also shown to exhibit superconductivity (Fronzi et al. Adv. Theor. Simul. 2020).
Myelofibrosis drugs and colorectal cancer markers
Working with CSIRO, Mt Sinai School of Medicine we have developed novel peptide antagonists of the thrombopoietin receptor that represent advanced leads for the first disease modifying treatment for the blood cancer myelofibrosis (Wang et al. Blood, 2016). We are also working with Monash University on discovering markers for colorectal cancer from lipidomic data using sparse feature selection and machine learning.
Antiviral drugs for coronavirus treatment
We are using very large-scale computational resources from Oracle Cloud Services to conduct molecular docking and molecular dynamics studies to discover drugs that can be repurposed for treating coronavirus infections such as COVID-19 (Piplani et al. PNAS, 2020).
Dr Monika Szabo Dr Mark Richardson Dr Shelvin Chand (CSIRO) Dr Marco Fronzi (UTS)
Collaborators
Prof Morgan Alexander Dr Grazziela Figieredo Prof Ricky Wildman (Nottingham) Prof. Alex Tropsha Dr Oles Isayev (UNC Chapel Hill) Dr Ira Katz Dr Geraldine Farjot (Air Liquide Santé International) Prof Nikolai Petrovsky Dr Sakhi Piplani Dr Puneet Singh (Vaxine, Flinders) Prof Joe Shapter (UQ) Prof Mike Ford (UTS) Prof Amanda Ellis (Melbourne) Dr Tu Le Dr Nas Mefati (RMIT) Prof Paul Pigram (La Trobe) Wil Gardner (La Trobe) Dr Aaron Thornton (CSIRO) Prof Nico Voelcker Dr Dave Rudd (Monash) Prof Michael Morris (Sydney)
Patents
TW9568/AU/PRV01; Anti-fibrotic compositions and devices comprising same - draft specification filed 26 May 2019
