Related Experiment Video
Updated: Aug 8, 2025
10:36
Rare Event Detection Using Error-corrected DNA and RNA Sequencing
Published on: August 3, 2018
12.2K
DeltaMSI: artificial intelligence-based modeling of microsatellite instability scoring on next-generation sequencing
Koen Swaerts1,2, Franceska Dedeurwaerdere3, Dieter De Smet1,2
1Department of Laboratory Medicine, AZ Delta General Hospital, Deltalaan 1, 8800, Roeselare, Belgium.
BMC Bioinformatics
|March 1, 2023
Summary
This study introduces DeltaMSI, a machine learning script for accurate DNA mismatch repair deficiency (dMMR) detection in tumors using next-generation sequencing data. DeltaMSI offers robust and high-throughput microsatellite instability (MSI) screening across all tumor types.
Area of Science:
- Oncology
- Bioinformatics
- Genomics
Background:
- DNA mismatch repair deficiency (dMMR) is critical for identifying microsatellite unstable (MSI) tumors.
- Current MSI detection methods, visual inspection or simplified bioinformatic scoring, are time-consuming or lack depth.
- Next-generation sequencing (NGS) data offers complex indel distributions for improved MSI analysis.
Purpose of the Study:
- To develop a machine learning script for robust dMMR screening using NGS data.
- To process the full complexity of indel distributions for accurate MSI detection.
- To enable high-throughput screening of dMMR in clinical tumor samples without paired normal tissue.
Main Methods:
- Trained 7 machine learning models using scikit-learn on normalized read depth data of 36 microsatellite loci.
- Integrated top logistic regression and support vector machine models into the DeltaMSI script for combined prediction.
- Evaluated DeltaMSI's diagnostic performance and robustness on diverse cohorts and sequencing chemistries, comparing it to mSINGS.
Main Results:
- DeltaMSI demonstrated superior robustness and diagnostic power (AUC=0.950) compared to mSINGS (AUC=0.876).
- Achieved 90% sensitivity at 100% specificity, highlighting clinical potential for high-throughput MSI screening.
- Validated in a real-world setting with 1072 unselected solid tumor samples.
Conclusions:
- DeltaMSI provides a robust and accurate method for MSI detection in clinical tumor samples.
- The script effectively utilizes machine learning to analyze complex indel distributions from NGS data.
- DeltaMSI is suitable for high-throughput screening of dMMR across all tumor types.
