Researchers at the University of Miami Miller School of Medicine research provided strong evidence that a new platform is a viable, accurate approach to mapping cell development.

Dr. Leor Weinberger, chair and professor in the Miller School’s Department of Cell and Systems Biology, and Binyamin Zuckerman, Ph.D., performed single-cell sequencing on C elegans embryos extracted in cold temperatures. The sequencing produced an enriched expression of the very early stages of embryonic development. The researchers then fed their data into CONCORD, a computational algorithm developed by researchers at the University of California, San Francisco.

CONCORD uses machine learning to create clean, clear maps that show how cells relate to each other in single-cell sequencing datasets involving development, cancer, the immune system and more. The system helps researchers clearly interpret high-complexity sequencing data and understand how cells behave, an integral element in creating effective therapies for diseased cells.

Testing the Cell Mapping Algorithm

“When we provided the raw data to CONCORD, it was completely blind. It’s basically a huge matrix of genes by cells,” Dr. Zuckerman said, who then produced a dimensionality reduction plot describing the heterogeneity of cell types present throughout the developmental process. “The ability to simplify gene expression data without losing the resolution of subtle differences between cells, and even preserving the time dimension representing embryonic age of the single cells is a revolution for the biologist seeking to interpret results in a meaningful way. You can actually see on the CONCORD map the first cell divisions. You can see that there are one-cell embryo cells, and then they split in two. And then the two cells split into four.”

CONCORD accurately reconstructed the lineage of the C. elegans embroyogenesis process. That CONCORD map based on Dr. Weinberger’s and Dr. Zuckerman’s RNA sequencing represents a startling advance in the science of single-cell data analysis.

“Usually when you look at dimensionality reduction plots, you see clusters of cells. But to make biological sense out of it is very difficult,” Dr. Zuckerman explained. “You need to try to figure out what’s happening with these clusters because the visual representation of dimensionality reduction plots does not provide a lot of intuitive information. Here, the dimensionality reduction is so good, it’s self-explanatory. You can actually see how the trajectory of embryogenesis happens.”

The Need for CONCORD

When labs study cells one at a time, technical differences like the machine used or the sample preparation protocol can make cells look different, even when they’re not. On top of that, many cells are in the process of changing from one type to another, which makes it hard to interpret their gene expression profile. Older tools sometimes erase real biology or invent false patterns.

CONCORD tries to keep what is biologically real while removing the noise. It looks at each cell twice, with small, random differences added, and uses machine learning to recognize that these two views belong to the same cell. At the same time, it compares that cell to others and learns how they differ.

What makes CONCORD special is how it chooses cells to compare. The system groups cells from the same dataset together, so the computer doesn’t confuse technical differences with true biological differences. And it intentionally compares cells that are very similar to detect subtle transitions, like when a cell is halfway through maturing into another type.

Why This is Important for Medicine

Researchers rely on massive, single‑cell datasets. CONCORD makes these datasets easier to combine, compare and interpret. That means more accurate identification of disease‑related cell types, a better understanding of how cells transition into harmful states, more reliable insights from multi‑center clinical studies and improved ability to find drug targets or biomarkers.

“This approach is superior to other methods, and our dataset is a powerful way to show the advantages of this analysis method,” Dr. Zuckerman said. “It’s a great validation. The results are amazing.”

By removing noise while preserving important biological signals, CONCORD helps scientists understand development, disease and treatment responses in a much more detailed and trustworthy way.

Source – University of Miami Miller School of Medicine