Researchers have developed a multitasking deep neural network that has been used to efficiently analyse a large dataset of specific immune cells from dengue patients in India1. This technique, designed to classify many single-cell data, was used to identify signatures of acute dengue infection.
The researchers say that this technique may help overcome the challenges of analysing single-cell data of many samples, containing millions of different cells. This will facilitate the proper diagnosis of various diseases.
Previous studies had shown the potential of neural networks in handling data based on heterogeneous clusters of cells.
An international research team, including a scientist from the National Institute of Mental Health and Neurosciences in Bangalore, India, developed SAUCIE, a deep neural network that could analyse single-cell data.
The researchers applied SAUCIE to 180 samples containing 11 million T cells, a type of white blood cells, from dengue patients and healthy individuals in India. They found 20 different clusters in the T cells. Five of them were CD8 T cells and 13 were CD4 T cells. It identified and clustered an important but rare group of cells that were generated due to an early immune response to dengue infection.
Besides, it indicated the differences between acute, convalescent and healthy subjects. The new technique also identified and clustered various data based on renal cancer cells, breast tumours, retinal bipolar cells and mouse brain more efficiently than existing methods.
Faster than the other methods, it presents a new way of using neural networks in the analyses of biological and biomedical data.
References
1. Amodio, M. et al. Exploring single-cell data with deep multitasking neural networks. Nat. Methods. (2019) Doi: 10.1038/s41592-019-0576-7
