British scientists believe algorithms can be used to make predictions by reading the biological language of cancer as well as with neurodegenerative diseases such as Alzheimer’s. This is based on utilizing big data collected through decades of research.
University of Cambridge scientists used the amassed data and inputted this into a computer language model. The aim was to assess if artificial intelligence is capable of detecting, predicting, and hence making earlier and more advanced discoveries. The AI was assessed in relation to experienced medical practitioners.
The starting basis of the research rests on the human body being home to thousands and thousands of proteins. A vast number of these have purposes that have yet to be identified. The aim is to look for subtle changes in these protein using a neural network-based language model. The researchers think that by learning the language of proteins then the mysteries of disease can be discovered.
One of the initial focal points is with shapeshifting biomolecular condensates. These are droplets of proteins found in cells . The researchers assessed that interpreting these provides the key for unlocking the language of biological function, and specifically the malfunctions that cause cancer and conditions like Alzheimer’s disease.
It is reasoned that some disordered proteins can form liquid-like droplets of proteins that do not possess a membrane. This enables the condensates to merge with each other. These structures can form and reform.
What is significant is that these complexes within cells are thought to affect gene expression (the way DNA is converted into proteins) as well as the mechanism by which cells make proteins (protein synthesis). Defects with this process are probably the causes of cancer and other diseases. The new algorithm aims to assess these events early and to provide new insights.
The development has been published in the Proceedings of the National Academy of Sciences, where the research paper is titled “Learning the molecular grammar of protein condensates from sequence determinants and embeddings.”
