How AI found the words to kill cancer cells

Researchers from UC San Francisco (UCSF) and IBM Research have created a virtual molecular library of thousands of "command sentences" for cells based on combinations of "words" that directed engineered immune cells to seek out and relentlessly kill cancer cells using new machine learning techniques.

The research, which will be published online on December 8, 2022, in Science, marks the first time that highly sophisticated computational methods have been used in a field that has up until now advanced primarily through ad hoc tinkering and engineering cells using pre-existing molecules rather than newly synthesized ones.

The development enables scientists to forecast which components, whether organic or synthetic, should be present in a cell to give it the specific behaviors necessary to react to complicated illnesses.

Wendell Lim, Ph.D., the Byers Distinguished Professor of Cellular and Molecular Pharmacology and the study's principal investigator, also serves as the director of the UCSF Cell Design Institute. We won't be able to create novel cellular treatments quickly that do the necessary actions unless we have such predictive capacity.

Get to know the molecules that form the words in cellular command phrases.

The process of selecting or developing receptors that, when added to the cell, will allow it to perform a new function is a significant part of therapeutic cell engineering. Receptors are molecules that span the cell membrane to perceive the surrounding environment and provide the cell guidance on how to react to it.

A kind of immune cell known as a T cell may be reprogrammed to detect and eradicate cancer cells by giving it the proper receptor. Some tumors have responded well to these so-called chimeric antigen receptors (CARs), but not others.

Kyle Daniels, Ph.D., a researcher in Lim's lab, who is the primary author on the paper, and Lim concentrated on the region of a receptor within the cell that contains strings of amino acids known as motifs. Each motif serves as an internal "word" of instruction that commands an action. What directives are carried out by the cell depend on how these words are put together to form a "sentence."

Today's CAR-T cells have been designed with receptors that tell them to attack cancer but also to rest after a while, kind of like telling them to "Kill some renegade cells and then take a breather." The tumors may thus develop further as a consequence.

The group thought they could create a receptor by mixing these "words" in various ways, allowing the CAR-T cells to complete the task without stopping. They created a library of roughly 2,400 randomly assembled command words and examined the potency of hundreds of them on T cells.

What can be learned about disease treatment from the language of cellular instructions

Then, Daniels collaborated with Simone Bianco, Ph.D., a computational biologist who was a research manager at IBM Almaden Research Center throughout the study's execution and is now the Director of Computational Biology at Altos Labs. Researchers Sara Capponi, Ph.D., also at IBM Almeden, and Shangying Wang, Ph.D., then a postdoc at IBM and now at Altos Labs, worked with Bianco and his colleagues to design whole new receptor phrases that they projected would be more successful using revolutionary machine learning techniques.

According to Daniels, "We modified some of the sentence's phrases and gave it a new meaning." The revised phrase instructed the T cells to "Knock those renegade tumor cells out, and stay at it" so that they could destroy cancer without stopping.

An innovative new research paradigm is produced by combining cellular engineering with machine learning.

The whole is unquestionably more than the sum of its parts, according to Bianco. It enables us to get a deeper understanding of the principles governing life itself and how living things function, in addition to how to create cell treatments.

Capponi said, "We will apply this method to a varied collection of experimental data and potentially reinvent T-cell design," given the success of the study.

According to the researchers, this strategy will provide cell treatments for regenerative medicine, autoimmunity, and other uses. Daniels wants to develop self-renewing stem cells so that there is no longer a need for blood donations.

According to him, the actual strength of the computational technique goes beyond creating command words and includes comprehending the molecular instructions' language.

According to Daniels, "it is the secret to producing cell treatments that accomplish precisely what we want them to do. This strategy "enables the transition from comprehending the science to engineering its practical implementation."