An international research team has developed a new strategy that can predict the potential clinical implications of new therapeutic compounds based on simple cellular responses.
This discovery was partly led by scientists affiliated with Université de Montréal (UdeM), and represents a major step forward in developing more effective drugs with fewer side effects, much faster than before.
The researchers conducted their work at Centre de Recherche de l’Hôpital Ste-Justine and published their findings in the prestigious journal Nature Communications.
Developing new drugs is a long, complex and costly process.
It starts with identifying the molecule or “ligand” (such as a drug, hormone or neurotransmitter) that can activate or block the target or “receptor” involved in a disease. Compound identification and validation is one of the most important steps in ensuring that a new drug provides an effective clinical response with the fewest possible side effects.
“Most new drugs tested on human subjects fail in clinical trials because the therapeutic response is insufficient. Developing a strategy that infers potential clinical responses early in the drug discovery process would significantly improve drug candidate selection,” said Besma Benredjem, the study’s co-lead author and a doctoral student in pharmacology at UdeM.
Finding the needle in a haystack
“Our main goal was finding a way to categorize a large number of drug candidates based on similarities in their effectiveness in triggering a multiplicity of cellular responses that help identify the therapeutic action of new compounds,” said Professor Graciela Piñeyro, co-senior author of the study and a researcher at CHU Sainte-Justine. To accomplish this, she worked with Dr. Olivier Lichtarge of Baylor College of Medicine, who uses advanced bioinformatic analysis to compare and group ligands according to fairly comprehensive signalling profiles.
Drugs produce desired or undesired clinical actions by changing basic signals within cells. By grouping drugs with known clinical actions and new ligands, we can infer the clinical actions of new compounds by comparing the similarities and differences in their signals with known drugs to promote desired clinical responses and avoid side effects.
This method of analysis was developed by using opioid analgesics as prototypes.
This made it possible for the team to associate simple cellular signals produced by opioids such as oxycodone, morphine and fentanyl with the frequency with which respiratory depression and other undesirable side effects of these drugs were reported to the Food and Drug Administration’s pharmacovigilance program.
At the height of the opioid epidemic, when the risk of death by respiratory depression is at its highest, the team believes this new analytical strategy could lead to the development of safer opioids.
“Thanks to our findings, we can now classify a large number of compounds while taking a multitude of cellular signals into account.
The wealth of comparisons this provides increases this classification’s predictive value for clinical responses,” said Professor Michel Bouvier, the study’s co-senior author and a principal investigator of molecular pharmacology and Chief Executive Officer of UdeM’s Institute for Research in Immunology and Cancer. “We think we can help patients by speeding up the drug discovery process so clinical trials can start earlier.”
“Our next goal is to use a similar approach to test cannabis products that may produce harmful neuropsychiatric actions among young people, and identify which cannabis extracts are most effective at treating chronic pain,” added Besma Benredjem.
In drug development field, novel drug candidates are being continuously discovered, yet the approval rate of the new drug is decreasing compared to the budgets spent on the R&D [1].
Due to the large chemical space, it is laborious and difficult to find new therapeutic compounds. Even if we find new drug candidates, most of them are filtered out in various screening steps, such as bioactivity and toxicity.
To solve this problem, many research has turned their attention to narrower chemical pools, like metabolites or natural products [2, 3].
When developing a new drug from a therapeutic compound, comparing its structure to known human endogenous metabolites is an essential step [4].
By comparing its structure, we may discover new positive effects or unwanted side effects caused by having a similar chemical structure of endogenous metabolites.
Some of the drugs have been developed to mimic the structure of human metabolites [5].
Antihistamine is one straightforward example.
By having the same binding site structure as the histamine with different residual structures, antihistamine antagonistically binds to histamine receptors and prevents histamine from initiating allergic reaction [6].
Other drugs like protirelin, which is a synthetic analogue of the thyrotropin-releasing hormone, or prednisone, which is an anti-inflammatory glucocorticoid derived from cortisone, have been synthesized to mimic human metabolites and bind to their target proteins [7].
As of now, DrugBank lists 2278 approved drugs, and Human Metabolome Database (HMDB) lists 29,266 recognized human metabolites [8, 9].
Among the 2278 approved drugs, 177 of them are exacted or derived compounds of human metabolites, according to the data in HMDB. While among 29,266 human metabolites, only 234 were developed into drugs. Such different proportions in drugs and human metabolites show the potential of new drug development from human metabolites.
In this study, natural products have been investigated to narrow down the chemical search space and find chemicals with high bioactivity. Natural products, which are secondary metabolites extracted from living organisms, have a distinct advantage in drug screening steps, for they are bioactive compounds in other organisms [2].
In previous work, principal component analysis (PCA) based chemical map was used to find natural products which are closely located to approved drugs in various feature spaces [10]. As a result, they found that natural products neighboring close to approved drugs show the same biological activity.
By definition, human metabolites also fall into the category of natural products in a broad sense.
In the previous study, it has been shown that the natural products are more structurally similar to human metabolites than drugs are to human metabolites [11].
Also, the number of similar human metabolites of natural products were much larger than that of drugs. The result is relevant to the fact that natural products and human metabolites are both secondary metabolites in living organisms.
By combining both human metabolite and natural product information, we could narrow down chemical search space with expected biological activities.
Here, we present a systematic method to discover new therapeutic compounds from natural products by using human metabolite information.
Our main hypothesis is that natural products which are similar to human metabolites will have similar endogenous effects in our body. To define the similarity between natural products and human metabolites, we considered molecular and phenotypic properties by utilizing structure, target and phenotype similarities.
By encompassing these three different aspects of molecular interaction, we expect to capture major features representing similarity between natural products and human metabolites.
Using the three similarity features, we trained a support vector machine (SVM) model to match natural products to their similar human metabolites and assigned verified phenotype terms to each pair. From the result, we expect to find possible therapeutic effects of natural products, which may serve as new leads to drug developments.
More information: Besma Benredjem et al. Exploring use of unsupervised clustering to associate signaling profiles of GPCR ligands to clinical response, Nature Communications (2019). DOI: 10.1038/s41467-019-11875-6
Journal information: Nature Communications
Provided by University of Montreal
