Drug Discovery for Antisense Oligonucleotides
- Reference number
- SM20-0048
- Start and end dates
- 210101-240630
- Amount granted
- 1 379 897 SEK
- Administrative organization
- Göteborg University
- Research area
- Computational Sciences and Applied Mathematics
Summary
We will develop a Digital Twin of Oligonucleotide Hybridization for predicting potency and safety of antisense oligonucleotide (ASOs) drugs. We will leverage existing knowledge from related, well-studied molecular reactions by developing machine learning models predicting reaction energy in this setting, and transferring models to the chemically modified ASOs. With active learning we will minimize the experimental effort necessary for increasing fidelity of predictions to the level required for ASO drug development. For drug safety we will assure with efficient ML algorithms that unintended hybridization reactions to any of the very large number of possible reactions partners in the transcriptome (the large part of genomes expressed as RNA) are unlikely to occur. Most importantly, we will use a novel pan-transcriptome representation of human genetic variation already in the in-silico design stage, which will lead to safer and more efficacious drugs, possibly on specific cohorts defined by their genetics. The expected results are joint high-impact publications and, at least at the prototype level, an implementation of computational methods transforming the speed of development of novel ASO drugs evaluated in on-going drug development processes at AstraZeneca.
Popular science description
Most classical drugs, e.g. Aspirin, are what is known as small molecule drugs. In the case of Aspirin the molecule consists of only nine Carbon, eight Hydrogen and four Oxygen atoms. The molecule acts by binding to proteins produced in the body needed for transmitting pain information. The small size of drug molecules often limits the specificity of the binding, which means that a drug might affect several processes in the body, might have unintended consequences, or interact easily with other drugs. This is one of the reasons why there are still no drugs available for some diseases despite very large and expensive efforts by the pharmaceutical industry. Antisense oligonucleotides (ASOs) are a new class of drugs which work very differently. Similar to the DNA and RNA molecules in our cells, oligonucleotides are simply chains of nucleotides, each nucleotide larger than the Aspirin molecule. DNA encodes our genes, and RNA is predominantly an intermediate representation of these genes which get translated to proteins. Through clever chemistry, ASOs are able to bind to specific RNA molecules and can for example cause protein production from them to be suppressed. ASOs can be very specific, as the recognition of targets is based on the sequence, e.g. “aattgtcata cgacttgcag tgagcgtcag gagcacgtcc aggaactcct…” for the protein Aspirin targets. The exchange aims at developing AI methods which allow to rapidly identify suitable ASO for a given target protein implicated in disease and to assure that the ASO does not bind to unintended targets. Binding here is not a qualitative—bind, doesn’t bind—property but rather a quantitative one, with a biophysical quantity, the so-called binding or hybridization energy measuring how well or badly an ASO binds to a target. A particular focus will be placed on genetic variation. Instead of working with one human reference genome, one very long sequence of A, C, G, and T, we will use a pan-genome, respectively pan-transcriptome, graph which will reflect the known genetic variation of humans. This increases the chance that ASO drugs will work for individuals of diverse genetic backgrounds (and both sexes) and reduces the risk of side-effects in specific subgroups. The project will help to define future research relevant to the pharmaceutical industry, build connections between the university and AstraZeneca and open new opportunities for the students in our programs.