COMMITTEE CHAIR: Dr. Noushin Ghaffari

TITLE: COMPUTATIONAL ANALYSIS AND STRUCTURAL PRIORITIZATION OF CANDIDATE BACTERIAL TRANSPOSABLE ELEMENTS FROM LARGE-SCALE GENOMIC DATASETS

ABSTRACT: Transposable elements are mobile DNA sequences that contribute to bacterial genome variation, gene disruption, and genome rearrangement. Large-scale computational tools can identify thousands of candidate insertion sequences and related transposable elements, but these outputs often require additional filtering before candidates can be interpreted biologically. This thesis develops a computational workflow for organizing, validating, and prioritizing candidate e-coli bacterial transposable elements from TIGER-derived genomic data. TIGER (Target/Integrative Genetic Element Retriever) is a tool, developed by Sandia National Laboratories, that discovers and maps mobile genetic elements, such as transposable elements and genomic islands, in bacterial and archaeal genomes with high accuracy. The workflow began with a large candidate dataset of 176,709 entries, discovered by TIGER for e-coli genomes, and a reduced working set of 35,451 candidates. Candidate records were linked back to TIGER metadata using contig names and coordinate information, allowing sequence entries to relate to genomic location, neighboring genes, and predicted element boundaries. MMseqs2 clustering was then used to group related candidate sequences under strict and relaxed similarity settings. BLAST-based screening and ISFinder comparisons were employed to evaluate whether candidate groups showed similarity to previously documented insertion sequences. ISFinder is a dedicated database for bacterial insertion sequences. Structural validation was also performed by examining terminal inverted repeats, direct repeats, candidate length, and sequence ambiguity. The analysis showed that candidate prioritization requires multiple types of evidence. Clustering helped organize large datasets into interpretable groups, while structural validation helped identify candidates with features expected of insertion sequences. However, BLAST and ISFinder comparisons also showed that lack of a local database hit alone is not sufficient evidence of novelty. Follow-up review indicated that local database construction, candidate sequence extraction, boundary selection, and external reference checks are important for correctly interpreting candidate status. In summary, this thesis generated a reproducible computational framework for reducing large transposable element candidate datasets into smaller, better-supported groups for biological follow-up. Rather than confirming novelty from a single method, the workflow combines similarity, structure, genomic context, and database comparison to prioritize candidates for future validation.

Keywords: Bacterial genomics, BLAST, direct repeats, insertion sequences, ISFinder, MMseqs2, structural validation, transposable elements.

Room Location: S.R. Collins Building, Room 111L