A Conceptual Framework for Human-AI Collaborative Genome Annotation

Li Maisie, Whan Alex, McNeil Meredith, Starns David, Irons Jessica, Andrew Sam, Suchecki Rad

Abstract:

Genome annotation is a process of identifying the locations, structures, and functions of genes and other elements from genome sequences. This process relies heavily on automated methods due to their ability to quickly process large amounts of data. However, there are still challenges and limitations to the existing automated methods for genome annotation, especially given the incomplete and erroneous genome data that can make the process error-prone. A current solution for improving genome annotation is to perform manual curation after automated genome annotation. Although this solution cannot scale to large amounts of genome data, it suggests that human-AI (Artificial Intelligence) collaboration may be a viable solution for addressing some limitations in genome annotation.
In this work, we propose a conceptual framework for Human-AI Collaborative Genome Annotation (HAICoGA) that leverages the strengths of both humans and AIs. We also develop guidelines to help humans and AIs collaborate effectively in HAICoGA. We conclude by posing open research questions for further study on supporting people in developing human-AI collaborative genome annotation workflows in the future.

External IDs:doi:10.1093/bib/bbaf377