Related Experiment Video
Updated: Jan 2, 2026
10:36
Rare Event Detection Using Error-corrected DNA and RNA Sequencing
Published on: August 3, 2018
12.5K
Coverage profile correction of shallow-depth circulating cell-free DNA sequencing via multidistance learning
Nicholas B Larson1, Melissa C Larson, Jie Na
1Department of Health Sciences Research, Mayo Clinic College of Medicine and Sciences, 200 1st Street SW, Rochester, MN 55901, USA, Larson.nicholas@mayo.edu.
Pacific Symposium on Biocomputing. Pacific Symposium on Biocomputing
|December 5, 2019
Summary
We developed a machine learning method to correct coverage variability in cell-free DNA sequencing for non-invasive prenatal screening. This approach improves accuracy in detecting fetal trisomies by reducing false positives and negatives.
Area of Science:
- Genomics
- Bioinformatics
- Computational Biology
Background:
- Shallow-depth whole-genome sequencing (WGS) of cell-free DNA (cfDNA) is crucial for non-invasive screening assays like liquid biopsies and non-invasive prenatal screening (NIPS).
- cfDNA WGS presents significant coverage variability, exceeding typical sources like GC content, which can compromise the accuracy of copy-number alteration and aneuploidy detection.
Purpose of the Study:
- To develop and evaluate an empirically-driven coverage correction strategy for cfDNA WGS using multi-distance learning.
- To improve within-sample coverage profile correction and enhance the accuracy of NIPS for fetal trisomies.
Main Methods:
- A weighted k-nearest neighbors-style machine learning method was trained on cfDNA WGS data from non-pregnant donors.
- The method leveraged prior annotation information and was applied to NIPS samples to assess coverage variability reduction.
- Performance was compared against a traditional regression-based correction method using GC content and mappability.
Main Results:
- The machine learning approach significantly reduced coverage profile variability in NIPS samples by 26.5-53.5% compared to the standard regression method.
- Combining annotation information with GC content and mappability improved performance over using either feature set alone.
- Improved discrimination of positive fetal trisomy cases was observed, enhancing NIPS assay performance.
Conclusions:
- Machine learning offers a powerful strategy for substantially improving cfDNA WGS coverage profile correction.
- This enhanced correction leads to more accurate downstream analyses, particularly for NIPS trisomy screening.
- The proposed method demonstrates potential to increase the reliability of non-invasive genomic screening assays.
