DREAMS: deep read-level error model for sequencing data applied to low-frequency variant calling and circulating...

Area of Science:

• Genomics
• Bioinformatics
• Cancer Research

Background:

• Circulating tumor DNA (ctDNA) detection via next-generation sequencing (NGS) shows promise for cancer identification.
• Distinguishing low tumor DNA signals from sequencing errors in plasma is a significant challenge.

Purpose of the Study:

• To introduce DREAMS (Deep Read-level Modelling of Sequencing-errors), a novel method for estimating individual read position error rates.
• To develop DREAMS-vc and DREAMS-cc, statistical approaches for variant calling and cancer detection, respectively, leveraging DREAMS.

Main Methods:

• Development of DREAMS for precise sequencing error rate estimation at the read level.
• Implementation of DREAMS-vc for variant calling and DREAMS-cc for cancer detection.
• Generation of deep targeted NGS data from matched tumor and plasma DNA of 85 colorectal cancer patients for validation.

Main Results:

• DREAMS accurately estimates sequencing error rates, crucial for low-frequency variant detection.
• DREAMS-vc demonstrates superior performance in variant calling compared to existing methods.
• DREAMS-cc achieves enhanced accuracy in detecting cancer from plasma DNA, outperforming state-of-the-art approaches.

Conclusions:

• DREAMS provides a robust framework for improving the accuracy of ctDNA analysis.
• The DREAMS approach significantly advances variant calling and cancer detection in liquid biopsies.
• This method holds potential for more reliable non-invasive cancer diagnosis and monitoring.