Doctoral colloquium - Jana Černíková (16.3.2026)
Monday 16.3.2026 at 13:10, Lecture room I 9
Jana Černíková:
Computational analysis of bacterial plasmids
Abstract:
Plasmids are small, circular, extrachromosomal DNA molecules commonly found in bacteria. They can be transferred between bacterial cells through horizontal gene transfer and often carry genes that confer antimicrobial resistance (AMR), making them an important focus in studies of antibiotic resistance.
The goal of our work is to develop new bioinformatics methods for plasmid detection and comparison, using techniques from machine learning and comparative genomics. We explore computational approaches for classifying plasmid sequences based on high-throughput sequencing data. Using k-mer profiles, various sequence-derived features, and homology-based log-odds scores, we train machine learning models to distinguish plasmid reads from chromosomal ones. Our dataset, consisting of multiple E. coli isolates, presents significant challenges due to class imbalance -- plasmid reads are markedly underrepresented.
To our knowledge, no existing software has been specifically developed for plasmid read classification. Our models outperform several previously proposed tools on this task; however, those tools were originally designed for the classification of longer contigs rather than short sequencing reads. Interestingly, a simple homology-based decision rule that we designed outperforms our neural network models.