Using AI for Targeted Epitope Prediction for Innovative Self-adjuvant Vaccine for Leukoencephalopathy

The project has developed a model that can assist in making a vaccine for Progressive multifocal leukoencephalopathy (PML). In this 8-week challenge, a collaborative team of 50 AI engineers from all around the world joined together.

The problem

Belyntic creates novel self-adjuvant vaccines based on synthetic peptides in the space of viral diseases. The focus lies on diseases with a clear unmet medical need. This Omdena-Belyntic project will focus on developing a model for a therapeutic vaccine against a deadly disease without a cure, called Progressive multifocal leukoencephalopathy (PML).

The primary characteristic of PML, a rare and frequently deadly viral illness, is multifocal progressive deterioration or inflammation of the brain’s white matter. The JC virus, which typically exists and is controlled by the immune system, is what causes it. Unless an individual has a compromised immune system, the JC virus is not harmful and attacks almost exclusively patients having a severely weakened immune system. In general, the first few months of PML have a death rate of 30–50%, and survivors may be left with varied degrees of neurological impairments.

At present, there are no medications that successfully limit or cure PML without causing harm. As a result, therapy seeks to reverse the immunological deficit in order to halt or stop the progression of the disease. This has given rise to the need to develop a vaccine for PML.

The project goals

The main goal of this project has been the accurate prediction of the best-suited epitope (the part of an antigen molecule to which an antibody attaches itself).

Project Scope:

• The collection of input data, i.e., DNA samples of patients with this rare condition, from different available sources has been completed.
• Standardization of the already collected data and definition of suitable reference sequences has been accomplished.
• Collection of more data from other sources has been conducted.
• Viral proteome and MHC class I and II binding affinity, obtained from different sources, have served as the prediction basis.
• Consideration of dependence factors such as different HLA types of patients and binding pockets has been taken into account.
• Prediction of the best-suited epitopes has been successfully carried out.