Related Concept Videos
Genomics
Genome Annotation and Assembly
Next-generation Sequencing
Next-Generation Sequencing Methods
Although all next-generation methods use different technologies, they all share a set of standard features....
You might also read
Related Articles
Articles linked to this work by shared authors, journal, and citation graph.
RAmpSim: a thermodynamic simulator for hybridization capture in metagenomic sequencing.
Primer Design through Submodular Function Estimation.
Accelerating String Comparison in RLZ Compressed Sequences via LCE Jumps.
Related Experiment Video
Updated: Sep 11, 2025

Author Spotlight: Advancing Alzheimer's Research – Exploring Early Detection and Multi-Omics Approaches
Published on: December 15, 2023
Tokenization and deep learning architectures in genomics: A comprehensive review.
Conrad Testagrose1, Christina Boucher1
1Department of Computer and Information Science and Engineering, University of Florida, Gainesville, FL, United States.
Advancements in DNA sequencing generate vast genomic data, necessitating better computational tools. Current deep learning tokenization methods for genomics are often inefficient or biologically inaccurate, requiring improved techniques for effective data analysis.
More Related Videos
Area of Science:
- Genomics
- Bioinformatics
- Computational Biology
Background:
- Modern DNA sequencing technologies have led to an exponential increase in genomic data.
- There is a growing demand for computational tools to analyze this data for applications like antimicrobial resistance and gene annotation.
Purpose of the Study:
- To survey current and foundational literature on deep learning architectures and tokenization techniques in genomics.
- To identify limitations in existing methods and suggest future directions for modeling DNA sequences.
Main Methods:
- Literature review of deep learning architectures and tokenization techniques in genomics.
- Analysis of the efficacy and biological relevance of current tokenization strategies.
- Comparison of sequence representation methods and their impact on scalability.
Main Results:
- Existing tokenization methods often struggle to capture or model underlying motifs in DNA sequences effectively.
- Many current tokenization approaches are either computationally inefficient, misrepresent biological motifs, or are adapted from Natural Language Processing (NLP) without sufficient biological consideration.
- Deep learning models are becoming more efficient, but tokenization remains a bottleneck.
Conclusions:
- Significant research is needed to develop efficient and biologically relevant tokenization techniques for genomic data.
- Future deep learning models in genomics should prioritize advanced tokenization strategies that accurately capture the information within DNA sequences.
- Improved tokenization is crucial for unlocking the full potential of genomic data analysis.

