Content overview : Since the 20th century, stem cell and regenerative medicine technology has been one of the hot frontiers in the international biomedical field. Now, researchers are beginning to explore turning stem cells into specific types of cells. However, during this process, stem cells will grow irregularly or differentiate into different types of cells spontaneously. Therefore, how to control the growth and differentiation of stem cells has become one of the challenges for researchers. In this paper, researchers from Zhao Yang's research group at Peking University and other researchers tried to apply machine learning to the process of pluripotent stem cell differentiation, which effectively improved the situation and brought a new direction for regenerative medicine.
Key words : pluripotent stem cells image analysis machine learning
This article was first published on the HyperAI super neural WeChat public platform~
Pluripotent stem cells (PSCs) are a type of pluripotent cells with self-renewal and self-replication capabilities, which can proliferate and differentiate into various cell types in vitro, replace damaged cells, and promote the recovery of damaged tissue functions . It has brought new hope for the treatment of eye diseases, cardiovascular system diseases and nervous system diseases.
However, the current process of directed differentiation of pluripotent stem cells has problems such as unstable line-to-line and batch-to-batch differentiation, which makes the preparation of functional cells time-consuming and laborious, and seriously hinders Research and development and large-scale manufacturing of pluripotent stem cell clinical application products. Therefore, it is particularly important to realize real-time monitoring of the differentiation process of pluripotent stem cells.
Recently, Zhao Yang's research group and Zhang Yu's research group at Peking University, together with Liu Yiyan's research group at Beijing Jiaotong University, developed a differentiation system based on bright-field dynamic images of living cells and machine learning, which can intelligently adjust and optimize the differentiation process of pluripotent stem cells in real time, and realize the Efficient and stable production of functional cells. At present, the research results have been published in the journal "Cell Discovery", titled "A live-cell image-based machine learning strategy for reducing variability in PSC differentiation systems".
The research results have been published in the journal Cell Discovery
Paper address:
https://www.nature.com/articles/s41421-023-00543-1
Experiment overview
Currently, microscopy techniques allow for image acquisition of cells, and machine learning methods allow for analysis of cell images. Therefore, this study uses machine learning algorithms to identify and classify cells in brightfield images to determine their lineages or cellular components, helping researchers to better understand cell structure and function.
It has been verified that the results of this study can effectively optimize and improve the differentiation process of pluripotent stem cells into cardiomyocytes (cardiomyocytes, CMs) and liver and kidney inlay cells. The whole set of research methods and procedures are as follows :
Figure 1: Machine learning optimized PSC-to-CM
a: The upper part of the figure shows that there is variability in each differentiation process of PSC, and the lower part of the figure shows that machine learning is applied to the above differentiation process, which effectively reduces the variability.
b: The PSC-to-CM differentiation process of the canonical Wnt signaling pathway is regulated by small molecule modulators. The green arrows indicate the duration and concentration of CHIR regulation in the first stage, and the colored dots indicate the checkpoints of machine learning.
c: Delayed brightfield images and cTnT fluorescence results over a 10-day period.
d: During the whole process, the location and morphology of successfully and unsuccessfully differentiated cells.
e: Texture and morphological changes of successfully differentiated cells from day 5 to day 12.
f: Line-to-line variability of differentiation efficiency.
g: Variability of different batches of cell differentiation.
h: Changes in local characteristics of differentiation images with different CHIR doses.
experiment procedure
Experimental dataset
Taking PSC-to-CM differentiation as the main example, the researchers used the Zeiss Cell Discover 7 live cell automatic imaging platform to collect bright field images in real time during the differentiation process and track the entire process, as shown in Figure 1b above. At the end of differentiation, successfully differentiated CMs were identified by fluorescent labeling with cTnT, a specific marker for cardiomyocytes. In this process, in order to increase the diversity of images, the researchers introduced several variables (different PSCs, initial cell density, differentiation medium, different CHIR doses), and finally collected more than 7.2 million images.
Experimental results
Combining live cell imaging technology and machine learning, this experiment achieved the following four results:
- Machine learning can accurately identify differentiated cell states and predict differentiation efficiency.
The researchers found that on day 6 of the differentiation process, the cells that could eventually successfully differentiate into CMs, i.e. CPCs (cardiac progenitor cells), began to take on a spindle shape,** so they used a weakly supervised model to identify such cells in brightfield images cells, and named them "Image-Recognized CPC (IR-CPC)". **As shown in Figure 2 below, the researchers concluded that the ratio of IR-CPC to total cells has a correlation with the true differentiation efficiency of 88%.
Figure 2: Correlation between IR-CPC ratio and true differentiation efficiency
At the same time, the researchers used the pix2pix deep learning model to predict the bright field images of the CM induction stage (ie, the first stage of differentiation), as shown in the figure below, and the correlation between the predicted differentiation efficiency and the real differentiation efficiency reached 93%.
Figure 3: Correlation of Predicted and True Differentiation Efficiencies
The above experiments show that machine learning can identify cell states at different stages of differentiation and can predict the outcome of differentiation in real time.
- Machine learning enables real-time prediction of differentiation time and inducer concentration.
During the differentiation process, the researchers found that the dose (concentration and treatment time) of the inducer CHIR99021 (CHIR) had a greater effect on the differentiation efficiency at the mesoderm stage (0-3 days). They constructed a logistic regression model based on the features related to CHIR in bright field images at the early stage of differentiation (0-12h) to predict the concentration of CHIR in wells (low, moderate, high). As shown in the figure below, in the selected CHIR When the treatment time was 24 hours, the accuracy rate of the model in judging the concentration of each well (a laboratory product with many small holes) reached 93.1%.
Figure 4: Model prediction of CHIR concentration in wells
At the same time, the researchers compared the prediction results (ie, deviation scores) of the models under different CHIR processing times (24h, 36h or 48h) to obtain the optimal CHIR processing time. As shown in Figure 5 below, the optimal CHIR processing time is about 12 hours (minimum deviation score). In addition, as shown in Figure 6, according to the prediction results of the model, the concentration of CHIR can also be adjusted to improve the differentiation efficiency.
Figure 5: Model predicts optimal CHIR processing time
Figure 6: Differentiation results with adjusted and unadjusted CHIR concentrations
The above experiments show that machine learning can achieve intervention in the dose of inducer.
- Machine learning can judge the optimal state of PSC initial differentiation in real time.
The researchers found that even at modest concentrations of CHIR, cells that failed to differentiate appeared, which they proposed was caused by spatially variable differentiation, whereby cells at the edge of the PSC colony at differentiation day 0 tended to succeed , while cells located in the center of the PSC colony are susceptible to failure.
In this regard, the researchers established a random forest-based machine learning model to identify the initial cell image features with a high differentiation success rate. The model results show that cells with medium cell area, longer edges and more potholes are easy to differentiate successfully. This is consistent with actual observations. Based on this model, the researchers found a 76% correlation between the identification of the initial PSC state predicted and the true differentiation efficiency, as shown in Figure 7 below.
Based on this, the researchers also changed the initial morphology of the cells through artificial intervention, effectively increasing the differentiation efficiency from 21.6% ± 2.7% to 88.8% ± 10.5%.
Figure 7: Correlation between identification of cell starting state and prediction of differentiation efficiency
The above results show that machine learning can perform quality control on the initial state of PSC.
- Machine learning can help screen small molecule compounds and improve differentiation stability.
The researchers found that CHIR concentration was one of the important factors affecting differentiation, so they tried small molecule screens to counteract inappropriate CHIR concentrations with new compounds. As shown in the figure below, the researchers built a small molecule screening platform based on the bright field live cell images on day 6 of the differentiation process and the established weak supervision model, and finally successfully screened BI-1347 among more than 3,000 small molecules. a compound.
Figure 8: The process of screening small molecule compounds by machine learning
The above experiments show that based on the machine learning model, researchers can build a small molecule screening platform, thereby shortening the screening experiment cycle and reducing screening costs, and the small molecules screened by this technology broaden the CHIR dose range, thereby improving the overall PSC Stability of the differentiation process.
Finally, in order to expand the application scenarios, the researchers applied the results of this study to the early stages of renal progenitor cells and liver cell differentiation, and also achieved accurate prediction results. It can be seen that the research results can provide real-time guidance for the differentiation process of pluripotent stem cells .
Cell therapy: Or a new track for biomedicine
Cell therapy is an emerging therapy that has shown promising results for a wide range of diseases (cancer, genetic disorders). Its main treatment methods are divided into immune cell therapy and stem cell therapy. Among them, stem cells have become one of the core research directions in this field due to their functions such as multi-directional differentiation, immune regulation, and secretion of cytokines.
At present, the development of cell therapy field in China is relatively short, but the future prospect is very broad . On the one hand, judging from the data, the next ten years may become a period of rapid growth in this field. According to relevant data, the market size of cell therapy in my country will increase from 1.3 billion yuan in 2021 to 58.4 billion yuan in 2030, with an average annual growth rate of 53%. According to other data, my country's cell and gene therapy market is expected to reach US$2.59 billion in 2025, growing at a compound growth rate of 276%.
** On the other hand, governments around the world have also continuously issued relevant policies to support and encourage this field. **For example, Beijing, Shanghai, Tianjin, Shenzhen and other places are vigorously developing the cell therapy industry. Shanghai launched the "Shanghai Action Plan for Promoting Cell Therapy Technology Innovation and Industrial Development (2022-2024)", proposing to strive for the scale of Shanghai's cell therapy industry to reach 10 billion yuan by 2024. Last year, Shenzhen successively issued documents to support the development of the biomedical industry, focusing on supporting the high-quality development of industrial clusters including cell therapy drugs.
Dataset and code address:
https://GitHub.com/zhaoyanglab/ML-for-psc-differentiation
Reference link:
[1]https://www.thepaper.cn/newsDetail_forward_23417694
[2]http://www.cls.edu.cn/Research/Research_Achievements6067.shtml
[3]https://stcsm.sh.gov.cn/zwgk/ghjh/20221104/f7b02ab5db40439e8d93f15b9dd206da.html
[4]http://legacy.frostchina.com/wp-content/uploads/2021/11/20211116-2.pdf
-- over--
This article was first published on the HyperAI super neural WeChat public platform~