NuProbe And Microsoft Researchers Invent A Deep Learning Model For Predicting NGS Sequencing Depth

Healthcare ManagementMachine LearningNews

By AIT News Desk On Jul 20, 2021

NuProbe Global, a genomics and molecular diagnostics company specialized in ultrasensitive sequencing assays, announced research presenting a deep learning model (DLM) for predicting NGS sequencing depth from DNA probe sequences. The findings, published in Nature Communications, are the latest research leveraging machine learning to improving the efficiency and accuracy of genomics. The study was co-authored by researchers from NuProbe USA, Rice University, and Microsoft Research UK.

Targeted DNA sequencing is the primary method for performing high-throughput genomics and clinical research and is increasingly adopted for molecular diagnostics. Differences in binding speeds between different genes and their corresponding DNA probes can result in non-uniform sequencing coverage, resulting in lowered clinical sensitivities and increased sequencing costs. Algorithms that can accurately predict DNA binding kinetics thus have the potential to improve the clinical impact of next-generation sequencing (NGS) panels.

“NGS panels with poor sequencing uniformity waste a large majority of sequencing reads and provide insufficient information on low-depth targets,” said Jinny Zhang, Senior Scientist at NuProbe USA and lead author of the study.

How ‘Unseen AI’ is Enhancing the Resilience of Critical Infrastructure

Nov 25, 2024

Efficient Query Optimization in Large-Scale AI Models

Nov 25, 2024

NetApp Joins the Vultr Cloud Alliance to Provide Scalable Data Management for Enterprise AI Workloads

Nov 25, 2024

Prev Next 1 of 40,986

“Our work shows that we can leverage neural networks and other advances in machine learning to help design NGS panels,” said Boyan Yordanov, co-author of the study and former Senior Scientist at Microsoft Research.

Data from the study highlights how the DLM can be used for predicting NGS sequencing depth for complex NGS panels containing over 30,000 probes. The DLM was able to predict NGS sequencing depth accurately with between 93% and 99% accuracy.

“This research result is part of our long-standing collaboration on applying machine learning to genomics,” said Andrew Phillips, Senior Principal Researcher at Microsoft Research UK, and co-corresponding author on the study. “Microsoft Research is committed to both basic and applied research at the intersection of computer science and biology.”

“Advances in NGS have enabled us as a society to adopt precision medicine, in which each patient’s personalized disease biomarkers are used to inform optimal treatments ,” said David Zhang, Head of Innovation at NuProbe USA, Associate Professor of Bioengineering at Rice University, and co-corresponding author of the study. “This collaborative research marks progress in computationally-designed NGS panels with better sensitivity that can lead to improved patient outcomes.