AI in ecogenomics

Developing insights into protein structure is fundamental as it directly dictates biological functions and interactions. Genetic variations have the potential to disrupt protein function.

In humans, these variations are of clinical importance, necessitating the identification of their associations with specific disease phenotypes.

In pathogens, variations can lead to resistance, underscoring the need to detect these changes to inform the design of effective drugs that prevent the spread of resistance.

See some of the avenues being explored by the Ascher Lab below.

Predicting effects of mutations on protein structure and function

Mutations in proteins can disrupt stability and interactions, leading to diseases such as cancer. Experimental methods to study these effects are often time-consuming and limited in scope, but advances in deep learning, especially graph-based and language models, offer new avenues to address these challenges.

The lab’s work in this area aims to leverage these deep learning techniques to predict how mutations affect protein stability and function. This is being used to inform clinical genomics efforts, and has been translated into the clinic for a number of genetic diseases already. It is also being used to rationally design new proteins for mining, bioremediation, biomanufacturing and as drugs.

Antimicrobial resistance and drug design

Studying antimicrobial resistance (AMR) is critical due to its role in confronting an escalating global health crisis. As pathogens increasingly become resistant to current treatments, the effectiveness of standard therapeutics has decreased, posing significant challenges in healthcare.

The Ascher Lab develops novel methodologies to understand and combat AMR which plays a key role in the design of drugs. In addition to detection and diagnosis of AMR, this is also providing insights to guide development of new antimicrobials and vaccines less prone to develop resistance (resistance resistant therapies).

Quantum machine learning for protein structures

Biological systems are complex, such as protein structures, their folding and interactions. In certain cases when dealing with high-dimensional data classical algorithms struggle in capturing insights from these systems. Quantum computers have recently demonstrated the promise of tackling problems that are currently intractable with classical computing architectures. With substantial experience in machine learning and deep learning, the Ascher Lab is actively developing quantum machine learning algorithms which can offer better insight into biological data.

Character-based structural tree for inferring deep evolutionary relationships

Protein structure is known to preserve evolutionary information over longer time-scales compared to protein sequence. Protein structure comparison can therefore uncover evolutionary relationships which may be difficult to recover using sequences. Early work on using structural characters has shown that more resolved phylogenetic analysis can be carried out using this new data modality as compared to the conventional use of structural distance data. With the ability to employ character-based evolutionary trees the Ascher Lab is working towards developing understanding of relationships between proteins, with potential applications beyond resolving the tree of life to areas like protein engineering.

The Ascher Lab is a strong advocate of open access research and therefore to facilitate access to the work carried out in the lab, web servers are developed to allow researchers the use of our databases and models.

AI in ecogenomics

Predicting effects of mutations on protein structure and function

Antimicrobial resistance and drug design

Quantum machine learning for protein structures

Character-based structural tree for inferring deep evolutionary relationships

Ascher Lab team structure