The mystery of how cells come together to form tissues and organs during the earliest stages of life has long puzzled developmental biologists. It’s like watching an intricate dance of cellular activity as cells shift, split, and grow in a carefully coordinated process. Into this fascinating dance, engineers at MIT saw a window —in embryonic development.
A team from MIT has made a pivotal contribution in this realm by creating a method to predict how individual cells behave during the earliest stage of fruit fly embryo development. This method, reported in a study published in Nature Methods, tracks cells on a minute-by-minute basis and could be a potential game-changer for understanding the development of more complex organisms. It might also prove instrumental in identifying early indication of diseases like asthma and cancer.
The engineers designed a deep-learning model that can analyze high-resolution videos of fruit fly embryos, each starting as a cluster of about 5,000 cells. This model predicts how each cell folds, divides, and reconfigures itself during the crucial first hour of embryonic development. The research team named this initial phase as ‘gastrulation’ where individual cells rearrange on a minute-by-minute scale. Ming Guo, associate professor of mechanical engineering at MIT and co-author of the study, noted that modeling this period enables intricate understanding of how local cell interactions contribute to the formation of global tissues and organisms.
Impressively, the model delivered 90% accuracy when predicting the dynamic behaviors of individual cells. This offers an unprecedented view into how a seemingly uniform embryo begins to develop unique structures. Alongside fruit flies, the team also sees potential in applying this model to other species, including zebrafish and mice, to identify universal patterns in embryonic development.
The innovative model can also prove critical for understanding the abnormal tissue formation associated with diseases. According to Haiqian Yang, MIT graduate student and co-author of the study, asthmatic tissues exhibit different cell dynamics when imaged live, and the model could capture these subtle dynamical differences. This comprehensive representation of tissue behavior will potentially improve diagnostics or drug-screening assays.
In a field where scientists traditionally model embryonic development as a point cloud, viewing each cell as a moving point, or as a foam, where cells are represented as sliding bubbles, Guo and Yang chose to combine the two models. By doing so, they were able to create a dual-graph structure that allows the model to track detailed properties such as the location of a cell’s nucleus, its interaction with neighboring cells, and whether it is folding or dividing at any given time.
For model training, the researchers used rare, high-resolution videos of fruit fly gastrulation provided by the University of Michigan. Then, using three of these videos for training, they tested the model on the fourth. The outcome was outstanding: the model accurately predicted not only the changes in each cell, but also the timing, down to the minute.
Despite the model’s readiness for wider application across other multicellular systems, including human tissues, the primary challenge lies in obtaining high-quality data. Guo expressed optimism, stating that if they could access quality data of specific tissues, this model could predict the development of many more structures.
Supported in part by the U.S. National Institutes of Health, this groundbreaking study is promising a revolution in biology and medicine. Find out more about it by reading the original article on MIT News.
This website uses cookies.