Approximately 100,000 unique proteins exist in the human body alone and many more exist in other organisms. These proteins perform several different functions that are vital for living organisims to function, thus understanding the function of each protein is extremely important. Many important proteins have been sequenced, however few have been annotated with their function since this is often a long and labourious process. This means there are several proteins that have been sequenced with unknown functions that could assist researchers.

Structured State Space Models (S4) described by Gu et al (2022) claim to outperform several other ML models when it comes to dealing with long sequential data. The S4 Sequence model has been shown to outperform Transformers and many other sequence modelling architectures on the Long range Arena benchmark tests indicating that performs well given sequences with +16 000 sequences. Since Protein sequences can go up to 35 000 amino acids that interact when folding, S4 is the ideal model to use to predict the function of protein sequences. This github repository consists of the code implemented in this paper.

• Organise files
• Sample of how to run
• Data
• License