Harvard Catalyst Profiles

While over 30 million individuals globally have undergone genome sequencing, the diagnostic yield for rare diseases using whole genome and exome sequencing remains below 25%. A significant barrier to improving this yield is the frequent oversight of many missense variants. Recent advancements in AI algorithms, such as AlphaMissense, have underscored the necessity for and challenges of such methods. However, the accuracy of these methods is not yet sufficient for deployment in clinical genetics. This project aims to harness advanced deep learning algorithms to aid in the interpretation of genetic variants. It involves analyzing extensive genomic datasets with detailed phenotype information to identify and interpret genetic variants and their potential impacts on various pediatric diseases. The project's scope includes comparative analysis of deep learning algorithms, development of ensemble approaches, and application to pediatric cohorts with genotype-phenotype data. The successful candidate will work under the guidance of a team of experts in computational biology, genetics, genomics, and deep learning. Required qualifications include prior experience or coursework in genetics, computational biology, or a related field, familiarity with deep learning techniques and tools, and proficiency in programming languages such as Python or R.