Learning about cancer cells using image recognition

At the most recent Moore-Sloan Research Seminar, Carlos Carmona-Fontaine explained how he is using image analysis to study cancer cell cooperation

Carlos Carmona-Fontaine at the Moore-Sloan Research Seminar (2017)

How can we stop cancer tumors from growing? One way is to discover conditions that accelerate their growth so that we can find ways to suppress precisely those conditions — and this is what Carlos Carmona-Fontaine from NYU’s Center for Genomics and Systems Biology is investigating.

At the most recent Moore-Sloan research lunch seminar, Carmona-Fontaine and his research team at the Carmofon Lab are studying how cells cooperate by placing them in unique 3D printed chambers called MEMICs (Metabolic Microenvironment Chamber).

MEMICs mirror the micro-environments of tumors, and allow biologists to test how cell samples react to specific variables like oxygen and CO2 levels, or varying degrees of proximity to blood vessels that provide key nutrients like glutamine. At present, Carmona-Fontaine’s team are combining MEMICs with high-throughput methods that allow them to incubate cells and follow the over time in roughly 800 different environments.

Their preliminary results so far suggest that tumor cell populations are especially susceptible to nutrient starvation. To overcome the problem, the lab noticed that these cells begin working together for collective survival.

But, as Carmona-Fontaine explained, this strategy is only effective when there is high cell density: low density populations die before they can cooperate for survival.

This is coined as the Allee Effect in biology — and it’s good news for us. If we can find a way to enhance the Allee Effect (e.g. keep cell populations low), cancer cells will not be able to survive under nutrient starvation even if they engage their cooperative strategies.

In the meantime, however, Carmona-Fontaine’s researchers still need to collect more data to support their preliminary results. As more cells continue to grow and die within the MEMICs, they are taking high resolution images of the cells so that the researchers can discern live cells from dead ones in each environment.

Did you know that GFP comes from a jellyfish called aequorea victoria?

With about 50,000 images to analyze, image recognition has made cell-counting easier — but only if the cells are stained with Green Florescent Protein (GFP) first. Staining cells is a time consuming process, so Carmona-Fontaine’s lab has just begun exploring how machine learning techniques can help them count cells without using GFP.

by Cherrie Kwok