Machine Learning

• Proteins


  Molecules that are some of the building blocks of life, essential for helping the body function
• Proteins


  • They are essential for helping us digest food, move and even think
  • They help build and repair the body's tissues
  • They keep the body working as it should
  • There are thousands of different proteins in the human body, each with one or more specific functions
• Protein drugs


  More and more proteins are now used in the development of drugs that treat illness and disease
• Finding a new protein for a drug


  1. Researchers look at thousands of proteins found in nature until they find one that comes close to doing what they want it to do

  2. Then comes the long and difficult process of getting the protein to do what they want it to in the body without any negative effects
• More than half the time the process of finding a new protein drug fails to deliver a protein that works
• Amino acids


  Molecular building blocks that make up proteins
• There are 20 different types of amino acids that make up most proteins
• The order of the amino acids determines the 3-dimensional structure of a protein, which determines the protein's function in the body
• Determining the order of amino acids coded by human genes used to be a lot of work, but thanks to the Human Genome Project, researchers are now able to do it very quickly
• Predicting how the chains of amino acids fold up and carry out a protein's function is difficult because there are so many ways a protein can fold, and it takes a huge amount of computing power
• Artificial Intelligence (AI) and Machine Learning (ML)


  Can help predict the ways a protein could fold
• Tools using ML to predict protein structures

  • AlphaFold
  • RoseTTAFold
• It takes a very large amount of data about proteins to help design new protein drugs, mostly collected from lab tests and clinical studies on patients
• Generative biology


  Using computers to learn from data to generate new data, such as designing protein drugs faster and with greater success, or designing proteins from scratch
• Predicting protein structure is not all that RoseTTAFold and AlphaFold can do, they can also model how proteins connect (bind) to each other, which is a key aspect of drug development
• Machine Learning could be used to create specific proteins that would bind to a specific target, which would be much faster than making them from scratch in the lab
• The hope is that Machine Learning can help figure out what proteins are useful in fewer steps and with fewer surprises, allowing drug companies to get treatments to people even faster than before