Forecasting Protein Aggregation with an Improved Algorithm

A new, improved algorithm for studying protein aggregation could help biologics manufacturers design better-performing products with less experimental effort. The software, developed by scientists based in Barcelona, offers the ability to analyze the aggregation of proteins drawn from the AlphaFold protein structure database, as well as helping companies identify more soluble alternatives.

“Protein aggregation is a bottleneck in the production and manufacturing of biologics,” explains Salvador Ventura, PhD, a professor in the department of biochemistry and molecular biology at the Autonomous University of Barcelona (UAB).

The problem, he explains, is that many proteins used as therapies evolved to be soluble at the concentrations found in the human body. But therapeutics, such as antibodies, are produced in as high a concentration as possible.

“We want the product to deliver the maximum dose with the minimum amount of injection,” he says. “But proteins aren’t designed to be soluble at these concentrations, and their aggregation causes different effects.”

These can include the patient’s immune system reacting negatively or the aggregated product ceasing to work.

To overcome this problem, Ventura says, companies and labs try to forecast protein aggregation, usually experimentally with high-throughput combinational assays. But these approaches are not convenient for startups or small spinoff companies.

A computational approach, such as his algorithm, now in its fourth generation, can help these companies predict and design around protein aggregation.

It offers the ability to draw protein structures from AlphaFold to analyze likely protein aggregation using simulations of molecular dynamics. Users, he says, can also choose to mutate selected parts of the protein, identify other proteins in the same family, and even look at the possible impact of pH on solubility.

“Our lab is both computational and experimental, so most of the designs we’ve made, we’ve already proved by experiment,” Ventura says.

Limitations include the scarcity of high-quality experimental data available to train the algorithm, he explains.

Going forward, the team intends to model which solution and formulation conditions best maintain the stability of therapeutic proteins in manufacturing and clinical settings. “We’re working on these next steps already,” he says. “Although, as yet, we don’t have an algorithm for this.”

Ventura spoke about the latest version of his algorithm at the Bioprocessing Summit Europe in March.

The post Forecasting Protein Aggregation with an Improved Algorithm appeared first on GEN – Genetic Engineering and Biotechnology News.