Machine learning tool for evolutionary Prediction

Proteins are the machinery driving the cell. To suit themselves to cellular demands, proteins change and evolve accordingly. This creates multiple variants of the protein, called alleles, with varying frequency in the population. Protein's evolution, as with the evolution of species, is affected by two main parameters: which sequence came first, and which sequence is best adapted to the protein's function and environment. A good example of this phenomenon is the HLA protein. HLA reports pathogens presence to the immune system cells, in order to activate and direct the immune response. The wide variety of pathogens urged the HLA gene to evolve and become the most diverse in the human genome. In an article from Frontiers in Immunology, Prof. Yoram Louzoun and his team describe a novel, machine-learning-based method to predict allele frequency of HLA. It is performed, for the first time, based on the sequence alone, yielding a natural measure for selection.