School of Bioscience
In this project we develop and implement an innovative Deep-Learning (DL) based method for quality assessment for industrial use. By using human embryonic stem cells as a model system, we develop in different steps a neural network (NN) classifier for a stepwise prediction of the cell state (quality) of these cells using quantitative PCR (qPCR) data.
The innovation and uniqueness of this project resides in two main aspects: interpretation of neural networks and transfer learning across technologies. Firstly, the interpretation of the representations learned in the initial NN classifier developed using single-cell RNA-seq data will allow the identification of transcriptomic signatures associated with cell quality. Secondly, the knowledge contained in the NN classifier developed using sequencing-based data will be transferred to a NN classifier that will use qPCR-based data as input.
By developing a NN classifier for quality control of cells based on qPCR data, we aim to develop an affordable method that can be implemented in the quality control process at our industrial partners, resulting in an increased accuracy for prediction of cell quality and the significant associated cost reduction. Likewise, the panel of biomarkers associated with cell quality identified by the DL-based model will serve as the basis for the development of sophisticated qPCR-based assays.
Finally, this method will also be integrated as package in dedicated advanced qPCR data analysis software for its commercialization.
The project is coordinated by the University of Skövde and executed in close collaboration with our industrial partners Takara Bio Europe Swedish filial, RISE, TATAA Biocenter AB, and MultiD Analyses AB.