1. Computational analysis of bacterial genomes and prediction of antimicrobial resistance. We develop computational tools for automated analysis of bacterial genome sequences. This includes assembly, validation of species, check and elimination of contamination by other species, determination of MLST type, building a phylogenetic tree from core genome sequences, detecting the presence and intactness of known resistance genes, detecting and describing the context of resistance genes, detection and description of plasmid sequences. Some of the tools for this analysis have been developed earlier in our workgroup, e.g. StrainSeeker (Roosaare et al., 2017), PlasmidSeeker (Roosaare et al., 2018) and PhenotypeSeeker (Aun et al., 2018).
Large-scale analysis of bacterial genomes can be done with different purposes. The most common aim is the epidemiological analysis of the spread of bacterial strains or their resistance genes (Telling et al., 2018; Bilozor et al., 2019; Sepp et al., 2019; Telling et al., 2020).
Second frequent aim of bacterial genome analysis is a functional analysis of genomic variants that appear under strong natural selection in a certain environment. A good example of this kind of analysis is the paper by Jõers et al., 2019, where we describe which genes are mutated in a strain that is less responsive to muropeptides in the environment and Aun et al., 2021, where we helped to detect the genes responsible for nodule colonization in Sinorhizobium meliloti.
Currently, our group is actively working towards developing better statistical prediction models that would predict virulence, antimicrobial resistance or other properties of bacterial strains, based on their sequence. For this, we use machine learning methods combined with biological knowledge.
Figure. Distribution of antimicrobial carbapenem resistance genes in bacterial isolates. Analysis performed and figure created by Grete Paat.
2. Development of PCR-based diagnostic tests. For more than 10 years we have been studying how DNA molecules hybridize to their specific and non-specific targets. This is important for designing highly specific PCR primers or microarray probes. We have written several software packages facilitating genomic PCR primer design. Our workgroup has participated in the development of widely used Primer3 software and we host an official webpage for Primer3, which serves approximately 150 000 users per month.
We have implemented our theoretical knowledge in practice in form of various DNA-based molecular tests. For example, we have developed diagnostic tests for respiratory disease pathogen detection for Estonian company Quattromed HTI (currently part of SynLab Eesti), blood sepsis-related pathogen tests for Dutch start-up company Microbiome Ltd (van den Brand et al., 2014) and food allergen detection test for Estonian biotech company Icosagen Ltd. We have developed multiplex PCR tests that would enable faster detection of MLST sequence types of food pathogen Listeria monocytogenes (Andreson, 2026).
Figure. Multilex-PCR experiment demonstrating the specificity of primers that we designed for each MLST type (groups of three). The universally present middle lane is signal from universal Listeria-specific primer pair used as a positive control. Data from Andreson et al., 2026.
3. Development of computational methods for the analysis of metagenomic data. A large advantage of DNA sequencing tests over PCR-based tests is that it can detect all pathogens simultaneously, unlike PCR that can detect only one pathogen per primer pair.
We are currently developing software and databases for the detection of pathogenic or allergenic species or strains from cell-free DNA from human blood samples.
Another line of current studies is food metagenomics. Good examples are papers Raime and Remm, 2018 and Raime et al., 2020. In these papers, we described methods and proof-of-principle tests for the detection of potentially allergenic food components using the metagenomic food sequencing approach.
Figure. Cookies with variable lupin flour content. After cooking at 200C, their DNA was extracted, sequenced and analyzed. Experiment performed and photo taken by Dr. Kairi Raime.
4. Development of fast and original computational methods for the analysis of personal genomes. Within the last ten years, we have developed several novels and fast methods for the detection of single nucleotide variants (SNV) from human personal genomes. FastGT (Pajuste et al., 2017) finds genotypes of all previously known SNV-s in a given personal genome in only 30 minutes. KATK finds all SNVs (known and novel from personal genomes in 3 hours (Kaplinski et al., 2021). In addition, we develop algorithms for detection of structural variants, that are frequently neglected in GWAS analyses. For example, we can detect polymorphic Alu element insertion sites (Puurand et al., 2019), gene copy numbers (Pajuste et al.,2023) and copy numbers of variable number tandem repeats e.g. those in front of ACAN, MAO, TRIB3 genes (Örd et al., 2020).
Our algorithms use a completely original approach to sequence analysis and this allows us to process human personal genomes ca 30x faster than traditional methods, without losing accuracy. Speed of the software is crucial in large scale analysis of individual genomes.
Figure. Typical distribution of frequency of unique k-mers wrt their depth of sequencing. Hypothetical distribution created by Tarmo Puurand and Märt Möls.