DREAMS

DREAMS models read-level sequencing errors from next-generation sequencing (NGS) data to improve low-frequency variant calling and circulating tumor DNA (ctDNA) detection.

Key Features:

• Deep read-level error modelling: Estimates error rates for each position within a sequencing read from NGS data to distinguish true genetic variants from sequencing errors.
• DREAMS-vc (Variant Calling): Incorporates individualized read-level error rate estimations into variant calling to enhance accuracy for low-frequency variants.
• DREAMS-cc (Cancer Detection): Uses refined read-level error models to improve identification and characterization of ctDNA in plasma samples.
• Performance on deep targeted NGS: Demonstrated superior performance compared with state-of-the-art methods for both variant calling and cancer detection using deep targeted NGS data from 85 colorectal cancer patients.

Scientific Applications:

• ctDNA detection in oncology: Improves sensitivity and specificity of ctDNA-based assays from plasma DNA for non-invasive cancer diagnostics and monitoring.
• Low-frequency variant detection: Enables reliable detection of low-frequency somatic variants relevant to tumor dynamics, treatment response, and disease progression.

Methodology:

Generates and analyzes deep targeted NGS data from tumor tissue and matched plasma samples and develops statistical models based on read-level sequencing error evaluation to distinguish true variants from sequencing artifacts.