Non-globular proteins in the era of Machine Learning

COST Action CA21160

THE ACTION

ABOUT US

The ML4NGP COST Action aims to establish an interdisciplinary pan-European network to address non-globular proteins, a fundamental research gap in the protein science field. this COST Action aims to describe the function and properties of Non-globular Proteins (NGPs) by combining experimental data and novel machine learning (ML) approaches.

Scientific Background

Protein structure prediction has long been considered the “Holy Grail” of structural biology. The recent success of AI-based tools (e.g. AlphaFold) has ushered in a new era of highly accurate structure prediction, bringing to light the secrets hidden in the three-dimensional structures of globular proteins, increasing our understanding about their structural features and molecular function. However, a large proportion of the proteomes from all domains of life are rich in sequences that do not fold into regular structures, commonly known as non-globular proteins (NGPs). NGPs comprise intrinsically disordered regions, repeats, low-complexity sequences, aggregation-prone and phase-separating sequences, and are implicated in a range of age-related diseases. These proteins have shaken the long-held structure-function paradigm where well-defined native protein structures are needed for function.

Their heterogeneous structural states and low sequence complexity challenge current experimental structure determination techniques and machine learning (ML) methods for structure prediction, making the molecular understanding of their sequence-structure-dynamics-function relationship difficult. The recent improvements of ML approaches and advances in determining NGP structural ensembles call for a timely re-assessment of the interplay between experiments and computation.

OUR VISION

The Action aims to strengthen both sides, in a way that machine learning can be used to enhance the combination of experimental methods and computational approaches to study NGPs. This Action aims that experimental frameworks are designed to provide information to computational methods, and computational methods are developed, trained and benchmarked with experimental data.

This integrative approach will provide new frameworks and methodologies to understand and to study and quantify the relationship between sequence-structure-dynamics-function of NGPs, stimulated by the development of new computational methods, standardization of best-practices for NGP experimental and computational research and the generation of novel curated datasets.

Action Details

CSO Approval date

27/05/2022

Start date

25/10/2022

End date

24/10/2026

Months

How to participate in the Action?

Inform the Action Chair of your interest (email).

Apply to join your Working Groups of interest.

Management Committee nominations are carried out through the COST National Contact Point.