Deep learning use case in genomics

Objective

• Choose a use case where machine learning (ML) technique is used to solve a genomics problem.

  • AlphaMissense - implements a deep learning model to predict the pathogenicity of DNA changes that alter protein sequence.

• Present a genomics scientist’s perspective of the ML workflow implemented in the use-case. i.e., genomics interpretation of the AlphaMissense deep learning workflow.

• Main focus is to highlight the importance of ML in solving genomics problems and not to promote AlphaMissense as a clinically viable product.

Content

• Genomics context of missense variants

  • Challenges in the current clinical setting

  • Current methods to predict the pathogenicity of missense variants

  • Ensembl Variation pathogenicity predictions

  • Limitations in current methods

  • Deep learning based solution for pathogenicity predictions

• Using deep learning models to predict pathogenicity

  • Why use deep learning?

  • Improving the accuracy of model predictions

• AlphaMissense

  • Studying genetic variations and disease outcomes

  • AlphaMissense deep learning workflow

  • Genomics context of each step deep learning workflow

• Data preparation

  • Data Preparation workflow

  • Dataset Splitting

• Feature engineering

• Model training and validation

• Model evaluation

• AlphaMissense community resource

• Hands-on experiment

Author/Instructor

• Pubudu Saneth Samarakoon