Detecting Random Mutations in 16S rRNA Sequences

As high-throughput sequencing technologies have surged, public sequence databases have experienced rapid expansion. Consequently, automated computational methods have become crucial for screening submitted sequences, with SILVA SSU Ref being a prime example. This database utilizes stringent algorithmic quality controls, yet it permits sequences to contain up to 30% deviation from any accepted sequence.

This laxity raises concerns about the potential contamination of public databases with modified sequences, such as those produced by DNA foundation models. Recognizing the need for methods to detect such sequences, researchers have investigated the detectability of altered 16S rRNA sequences. They have generated these sequences through random substitutions that meet the SILVA database's quality control criteria.

Employing classifiers based on conserved motifs, the researchers developed a classifier capable of distinguishing modified sequences from natural 16S rRNA. The best classifier achieved over 90% sensitivity and specificity when using a 5% artificial mutation rate. One key feature employed was gapped k-mers derived from universally conserved nucleotides, which were found to be conserved across all three domains of life, even though they relied on exact matches to patterns observed in E. coli.

This finding expands our understanding of conserved grammatical structure in small subunit rRNA sequences. The source code for the classifiers is now available at https://github.com/rainhaworth/16S-Mutation-Classifiers.