JERUSALEM, Jan. 7 (Xinhua) -- Israeli researchers developed a new framework designed to accurately label genomic data, which was a major challenge in biological research, the Hebrew University of Jerusalem (HU) said in a statement on Tuesday.
Genomic datasets often include samples that are labeled incorrectly or unclearly.
To address this issue, the team used artificial neural networks (ANNs) in a new way, as described in the journal Nature Computational Science.
Instead of just using the ANNs to make predictions, they focused on understanding which samples were harder for the networks to label correctly.
By studying the difficulty of labeling, the researchers found errors in cell annotations and improved the interpretation of biological data.
HU noted that this approach helps uncover key biological pathways related to development and disease and highlights a new way to handle single-cell and spatial omics data, which have revolutionized the ability to study cells in health and disease.
It added that the tool also provides a better way to analyze data from single cells and has great potential for advancing biological research.
In the long term, it could also improve how diseases are diagnosed and treated, according to the team.
