Using scalable proteomic tools for drug discovery

Gene expression profiling has long been the dominant method for determining how cells respond to drugs. Tools for this are fast, cheap, and scalable — especially compared to tools that measure protein levels or activity. However, profiles of gene expression may not fully capture all of the ways cells respond to drugs. In a new study published in Cell Systems, a team of researchers describes the feasibility of using cost-effective, scalable experiments to determine how novel drug candidates change the profiles of proteins within cells, thus providing essential clues for therapeutic research and development.

“This is really the first study which used protein profiling methods to investigate a large number of drugs — systematically recording their responses induced in cells,” said senior author Jacob Jaffe, an institute scientist and associate director of the Proteomics Platform at Broad Institute of MIT and Harvard. “What we are trying to do is generate a large library of biochemical responses elicited in response to drugs.”

All of the study's data are part of a publicly-accessible reference library called Touchstone-P, available in the Broad Institute's larger CLUE connectivity resource. Scientists can also query their own data against it.

Touchstone-P was inspired by the Broad Institute's Connectivity Map (CMap), which catalogs drugs' impacts on cells' gene expression. In the CMap context, drugs that elicit highly similar or highly dissimilar patterns of gene activity are said to be “connected,” giving scientists a valuable reference point for understanding biological connections between genes, diseases, and therapeutics, which in turn can inform clinical trials.

“We wanted to bring proteomic information into a similar framework as CMap,” Jaffe explained. “For the last five years, we have been developing high-throughput protein assays with the goal of making them scalable and to integrate our results with those of CMap’s gene expression profiles.”

“Connectivity helps you look at relationships between drugs or any sort of perturbations caused by them without necessarily focusing on the individual analytes that you measure in an assay,” said Lev Litichevskiy, lead author and associate computational biologist in Jaffe’s lab.

To build Touchstone-P and extend the CMap approach to proteomics, Jaffe and his colleagues used mass spectrometry and measured how drugs altered protein signaling pathways as well as modified epigenetic markers — each being essential biomarkers of a disease state. They tested 90 drugs across five cancer cell models (breast, lung, pancreatic, prostate, and skin cancer cell lines) and one neurodevelopmental cell model.

These complete matrices of connectivity data, collected through the quantified protein profiles, can be browsed and visualized in the reference library.

A heatmap and related data for a drug in the Touchstone-P library.
Credit: Lev Litichevskiy

The authors hope that with the additional knowledge from proteomic signatures, they will be able to provide fine-grained information about the effects of a drug before bringing it to clinical trials, thereby potentially saving tremendous cost.

“One of the amazing things about the connectivity concept is that once you are able to compare a molecular profile in an assay to a reference library of other molecular profiles, you can more easily integrate data across assays,” said Jaffe. “Connectivity breaks down a lot of barriers between the different types of assay data we get in drug discovery research.”

Future studies will involve expanding the library to include more drugs and more cell types, for instance, neurobiology and cardiovascular models. Jaffe said, “We will also be looking at combinations of drugs at a time to learn how their effects either impinge or enhance one another.”

This work was funded in part by the NIH Common Fund’s Library of Integrated Network-based Cellular Signatures (LINCS) program.