DNABERT, a linguistic approach for sequential predictions within Biology and Health

Kassab, Daniël (2024) DNABERT, a linguistic approach for sequential predictions within Biology and Health. Bachelor's Thesis, Biology.

Preview	Text bBIO_2024_KassabDB.pdf Download (385kB) | Preview
	Text toestemming.pdf Restricted to Registered users only Download (149kB)

Abstract

The prediction of cis-regulatory elements through many predictive models is widely used within Medical- and Biological research, but it is not perfect because their function differs depending on their context. Therefore, DNABERT is introduced to face this limitation by taking into account context during its predictions by considering the RNA- and DNA sequence as a language. In this literature thesis, I discuss whether this new model can achieve revolutionary performance compared to other existing models by comparing their accuracy scores for promoter- and splice site prediction. Furthermore, the mechanism of this new model was explored to provide the essential background information, which is needed for further customization and optimisation. This was performed by adapting the methods of DNABERT from Ji et al. 2021 and DNABERT-2 from Zhou et. al 2023. Consequently, an overview is provided of the promoter prediction mechanism by DNABERT using the described method by Ji et al. 2021. Finally, I identified the application possibilities for this new model in Medical- and Biological research. I conclude that this new DNABERT model should be the “first-choice” for performing DNA- and RNA predictions and that it should only be used as an additional tool. It should never be used as a replacement for decision-making within diagnostics and research. Finally, this model should still be further improved and customised to enlarge its impact on Health and Biology.

Item Type:	Thesis (Bachelor's Thesis)
Supervisor name:	Jong, A. de and Moll, G.N.
Degree programme:	Biology
Thesis type:	Bachelor's Thesis
Language:	English
Date Deposited:	18 Jan 2024 08:29
Last Modified:	18 Jan 2024 08:29
URI:	https://fse.studenttheses.ub.rug.nl/id/eprint/31828

Actions (login required)

View Item