Stem cells are still not well understood, but there is a lot of interest in knowing more about them for many reasons; they have tremendous therapeutic potential, they are critical players in development, and they have many unique characteristics. Scientists have now applied machine learning to hematopoietic stem cells to learn more about the decisions that lead to their fate as a more specialized blood cell. The researchers are featured in this video below explaining the work, which was published in Nature Methods.
"A hematopoietic stem cell's decision to become a certain cell type cannot be observed. At this time, it is only possible to verify the decision retrospectively with cell surface markers," explained Dr. Carsten Marr, head of the Quantitative Single Cell Dynamics Research Group at the Helmholtz Zentrum München's Institute of Computational Biology (ICB).
Using deep learning, which simulates human learning in a computational framework, Marr's team have created an algorithm that predicts stem cell fate decisions. "Deep Neural Networks play a major role in our method," said Marr. "Our algorithm classifies light microscopic images and videos of individual cells by comparing these data with past experience from the development of such cells. In this way, the algorithm 'learns' how certain cells behave."
A collaborator, Timm Schroeder at ETH Zurich, worked to film stem cells. With the data that was gathered on appearance and speed, the software could track and remember behavioral patterns, which led to the ability to make predictions. "Compared to conventional methods, such as fluorescent antibodies against certain surface proteins, we know how the cells will decide three cell generations earlier," said ICB scientist Dr. Felix Buggenthin, who was co-first author of the work, along with Dr. Florian Büttner.
Marr explained the potential applications and benefits of the work; "Since we now know which cells will develop in which way, we can isolate them earlier than before and examine how they differ at a molecular level. We want to use this information to understand how the choices are made for particular developmental traits."
The scientists are not limiting themselves to hematopoietic stem cells. "We are using Deep Learning for very different problems with sufficiently large data records," said Prof. Dr. Dr. Fabian Theis, ICB director and holder of the Mathematical Modeling of Biological Systems Chair at the TUM, who was study leader with Carsten Marr. "For example, we use very similar algorithms to analyze disease-associated patterns in the genome and identify biomarkers in clinical cell screens."