Artificial Intelligence Used to Gain Deeper Understanding of Cancer Biophysics

Artificial Intelligence Used to Gain Deeper Understanding of Cancer Biophysics

A team of researchers from the Tufts University’s School of Arts and Sciences, the Allen Discovery Center at Tufts, and the University of Maryland have developed an artificial intelligence-derived model that has been used to gain an insight into the biophysics of cancer.

AI systems are capable of human-like learning and can process vast amounts of data to identify patterns, predict outcomes, and arrive at conclusions that would take teams of humans many years. AI systems therefore have considerable potential for use in biomedicine. In biomedicine, there is certainly no shortage of data. The problem is accessing and processing all of those data. AI-based systems could be used to see patterns in data that would otherwise go unnoticed and to obtain results that would otherwise be unachievable.

In this case, such as system has been developed by the researchers, who used it to determine a complex biological process that would cause tadpoles to develop a cancer-like phenotype. The researchers had previously proved that it was possible to turn melanocyte pigment cells in tadpoles into a cancer-like form by disrupting bioelectrical and serotonergic signaling.

While the melanocytes could be converted to the cancer-like form, in the experiments the researchers noted that it was an all-or-nothing process. All melanocytes in the frog larva were converted or none were. There were no cases where only some of the melanocytes were concerted to cancer-like forms.

In order to determine how the cells were making ‘group decisions’, the researchers turned to their AI-derived model to reverse-engineer the process. The AI-derived model was used to run 576 virtual experiments. 575 of those experiments failed to produce the desired result, although the final test yielded an answer.

The model predicted three reagents – a 5HTR2 inhibitor called altanserin, a mRNA encoding constitutively active CREB – VP16-XlCreb1, and a VMAt inhibitor – that were capable of generating the cancer-like phenotype. The model also determined the exact concentrations of those reagents to produce the result.

Those reagents were then mixed in the relative concentrations and were used in an in vivo study on tadpoles. Instead of the all-or-nothing conversion to cancer-like phenotypes, only some of the cells converted: Something that the researchers had never before seen.

Finding the right combination of reagents to produce that effect could have taken the researchers years without the AI-model, if they discovered the correct mix at all. They had already spent years conducting a diverse range of experiments and had yet to see partial conversion of the melanocytes.

The AI-derived model also predicted the percentage of tadpoles that would remain completely normal to within 1% of the results of the in vivo study, as well as those that would display full or partial conversion.

First author of the study, Daniel Lobo, Ph.D., said “Even with the full model describing the exact mechanism that controls the system, a human scientist alone would not have been able to find the exact combination of drugs that would result in the desired outcome.” Lobo also explained the significance of the test, “This provides proof-of-concept of how an artificial intelligence system can help us find the exact interventions necessary to obtain a specific result.”

Leave a Reply