A Machine Learning Assisted, Label-free, Non-invasive Approach for Somatic Reprogramming in Induced Pluripotent Stem Cell Colony Formation Detection and Prediction

Abstract

During cellular reprogramming, the mesenchymal-to-epithelial transition is accompanied by changes in morphology, which occur prior to iPSC colony formation. The current approach for detecting morphological changes associated with reprogramming purely relies on human experiences, which involve intensive amounts of upfront training, human error with limited quality control and batch-to-batch variations. Here, we report a time-lapse-based bright-field imaging analysis system that allows us to implement a label-free, non-invasive approach to measure morphological dynamics. To automatically analyse and determine iPSC colony formation, a machine learning-based classification, segmentation, and statistical modelling system was developed to guide colony selection. The system can detect and monitor the earliest cellular texture changes after the induction of reprogramming in human somatic cells on day 7 from the 20–24 day process. Moreover, after determining the reprogramming process and iPSC colony formation quantitatively, a mathematical model was developed to statistically predict the best iPSC selection phase independent of any other resources. All the computational detection and prediction experiments were evaluated using a validation dataset, and biological verification was performed. These algorithm-detected colonies show no significant differences (Pearson Coefficient) in terms of their biological features compared to the manually processed colonies using standard molecular approaches.

The application of convolutional neural network to stem cell biology

“…In ghost cytometry, which is cell sorting without molecular labels, morphological features are converted to wave data using a random barcode system to classify and sort cells [53]. A machine learning algorithm can also be used to classify cell morphology [54,55], cardiac tissue contractility, and molecular imaging [56].…”

Stem cell imaging through convolutional neural networks: current issues and future directions in artificial intelligence technology

“…A lack of the necessary downstream differentiation into functional cells is possible if the colony identification is not consistent. Therefore, an automated quantitative methodology with stable or constant colony maturation identification is needed to effectively aid biologists during the iPSCs production stages (Fan et al, 2017).…”

“…Then, Fan et al (2017) reported on a time-lapse-based imaging study using bright-field microscopes that measured the morphological changes of the mesenchymal-to-epithelial transition during the cellular reprogramming that precedes iPSCs colony formation. However, the use of iPSCs in further applications may be limited due to the quality of iPSCs, which can only be checked by colony determination.…”

“…Recently, iPSCs have gained popularity because their usage does not raise major ethical issues, as ESCs do (Fan et al, 2017). Moreover, iPSCs are known to express specific genes and proteins as well as differentiation patterns that are believed to be identical to ESCs (Chagastelles & Nardi, 2011;Nordin, Ahmad & Farzaneh, 2017).…”

Figure 3: Convolutional neural network architecture.

“…ESCs and iPSCs are two forms of pluripotent stem cells (PSCs). The former derives from the early stage of an embryo, and the latter is obtained through a genetic reprogramming procedure in which terminally differentiated somatic cells are reversed back to the pluripotent state (Fan et al, 2017). Crucial morphological changes take place during PSCs differentiation.…”