A statistical framework for detecting mislabeled and contaminated samples using shallow-depth sequence data

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Replicate DNA sequencing is common for data validation.
• Errors like contamination or sample mix-ups can occur during sequencing.
• Existing error detection methods are often limited, especially for low-depth data.

Purpose of the Study:

• To develop a robust method for detecting errors in DNA sequencing replicates.
• To overcome limitations of existing ad hoc and pairwise comparison methods.
• To provide a tool suitable for various sequencing depths.

Main Methods:

• Utilized Bayes Theorem to calculate posterior probabilities.
• Inferred relationships between putative replicate samples.
• Developed an R package named BIGRED (Bayes Inferred Genotype Replicate Error Detector).

Main Results:

• The new method is suitable for shallow, moderate, and high-depth sequence data.
• Accurate error detection was achieved in simulation experiments.
• The approach can infer which samples originate from an identical genotypic source.

Conclusions:

• The BIGRED method effectively addresses limitations of current error detection approaches.
• It offers a reliable tool for ensuring data integrity in genomic studies.
• The R package is freely available for researchers.