Using AI to Find Peptide Therapeutics for COVID-19 | BioSpace
Originally published Apr 24, 2020
By now, I’m sure you know the breakneck pace researchers are moving to find drugs and develop a vaccine against SARS-CoV-2, the novel coronavirus that causes coronavirus disease 2019 (COVID-19).
a biotech company that focuses on leveraging artificial intelligence (AI) to discover therapeutic peptides, is joining the fight by tasking their powerful AI platform to find peptides that can be used as therapeutics for COVID-19.
“We apply AI to the largely untapped landscape of peptide chemistry to find peptide therapeutics,” Nora Khaldi, Ph.D., Nuritas CEO, Founder, and acting CSO, told BioSpace. “We aim to find bioactive, stable, cell-penetrating peptides with better bioavailability and activity against a target or pathway of interest.”
Peptides, short amino acid chains (think small proteins), are vastly abundant in every living thing. They are the “Goldilocks” of molecules because of their size — not too small (like small molecules) and not too big (like biologics such as antibodies). They are big enough to bind cellular targets with a large binding pocket but small enough to enter into the cell, giving them access to many more targets inside the cell.
“Peptides can go after targets that small molecules and biologics can’t, like unstructured targets and targets with a large binding pocket,” explained Khaldi.
Despite their unique abilities and vast customizability, peptides have been largely unexplored as potential therapeutics. As of March 2017, there were only 60 approved peptide drugs on the market, with 155 more in active clinical development.
“We wanted to focus on peptides because they are the best natural way that our body communicates,” Khaldi added. “All the solutions to our medical problems are out there — we just haven’t found them yet.”
Nuritas’ AI platform
Nuritas’ Peptide Finder (ΝΠΦ) Platform combines a database with years of experiments on thousands of peptides with state-of-the-art machine learning (ML) architecture that can identify peptides active against a target or pathway of interest.
“We use the results of thousands of experiments we’ve done over the past five years along with data from our proprietary proteomics database and very carefully curated data from the scientific literature as training data for our supervised and unsupervised ML models,” Khaldi explained.
In addition to their in-house testing, Nuritas collects proteome samples from around the world to characterize via mass spectrometry, allowing researchers to identify and quantify peptides of interest amongst the 30 billion peptides that are typically present in a single proteome source.
Nuritas also uses software that can “read” scientific literature published online, automatically extracting meaningful, relevant data to add to their database. Data curators then parse through the extracted information to ensure its quality before adding it to the body of peptide data fueling the platform.
To begin the peptide searching process, you input what you are looking for, much like checking a bunch of tick boxes or filtering your search in a search engine. Once you have selected the desired peptide properties and identified the target, the algorithm will search the database and find peptides matching your search query.
The algorithm can perform both targeted and non-targeted searches, depending on if you know the target of interest or not. If a target is unknown, a pathway that you want to modulate, like one dysregulated in a disease, can be input into the AI algorithm. AI will find peptides that can modulate the desired pathway, enabling researchers to then conduct experiments in the lab to determine what each peptide is targeting within that pathway.
Once peptide hits are found, they are synthesized and tested in the lab. The test results are put back into the database and the algorithm runs again with this new information, generating even better-matched peptide hits. This process is repeated a few times before promising peptides are identified.
Over 60 percent of the peptides identified by the algorithm have the desired activity against the target once they are made and tested in the lab.
“You reduce the risks of drug development by incorporating toxicity, biorelevance, and target data into the initial peptide search, so you know those peptide hits bind to the desired target with a suitable toxicity profile,” Khaldi commented.
Verified peptide hits can be further optimized via computer modeling (called computational chemistry) to add desirable characteristics, such as increased stability of the peptide. These optimized peptides can then be made and tested in clinical trials.
In partnership with BASF, the Peptide Finder Platform identified the first AI-discovered anti-inflammatory product, called PeptAIde, which is a plant-based peptide cocktail designed to regulate inflammation after exercise.
Using their AI platform for COVID-19 drug discovery
Now, Nuritas is expanding their discovery efforts to explore a treatment to address the symptoms of COVID-19.
“There is currently a lot of research working towards a COVID-19 vaccine, which is great, but we also need to focus on finding therapeutics to treat people with COVID-19,” Khaldi said. “Even with a vaccine, there will be people who will get sick and need medicine.”
Nuritas is using its AI platform to search for peptides active against COVID-19 targets. They will have two projects focusing on COVID-19 peptide identification: one for identifying antiviral peptides that target how SARS-CoV-2 hijacks cells, and one for identifying peptides that locally reduce lung inflammation without suppressing the entire immune system.
For the first project, Nuritas researchers will be searching through a database of peptide drugs that have already been tested in clinical trials. By repurposing peptides that have already been studied in the clinic, they can rapidly move any SARS-CoV-2-active peptides into clinical studies for COVID-19.
SARS-CoV-2, like other coronaviruses, infects human cells by using the ACE2 receptor (and a few other enzymes, like proteases), then hijacks the cells to churn out more copies of the virus. The first project aims to identify peptides that block the virus at any point, from initial entry into the cell through the viral replication process. Scientists can identify peptides that bind to known targets and simultaneously identify new COVID-19 targets that can be modulated by peptides.
“In general, viruses hijack cells via protein-protein interaction,” Khaldi explained. “How do we stop that interaction and, ultimately, the virus from hijacking the cell? The best way is by using a peptide.”
It will take 4–5 months to find peptide hits using the AI platform. Once promising peptides are found, they will be made or obtained from the company who originally made the peptide and passed along to collaborators who can test the peptides on COVID samples and tissues.
The second project is longer-term and aims to address the lung issues that coronavirus patients face while acutely sick and after recovering. A 15-year follow-up study on 78 severe acute respiratory syndrome (SARS) patients in China showed that lung fibrosis was a long-term issue.
Certain immune markers of inflammation, such as interleukin-6 (IL-6) and tumor necrosis factor-alpha (TNF-alpha), are significantly elevated in COVID-19 patients. IL-6 inhibitors approved for other diseases are already being tested in clinical trials with COVID-19 patients, including tocilizumab (Actemra) in China, the U.S., and other countries, and sarilumab (Kevzara) in the U.S. and other countries.
However, IL-6 inhibitor drugs induce systemic immunosuppression. Nuritas aims to create a cocktail of inhaled peptides that directly reduces inflammation in the lung to normal levels without suppressing the entire body’s immune system. Ideally, COVID-19 patients would start taking the inhaled drug at the beginning of their symptoms to reduce the amount of inflammation and lung fibrosis before it gets worse.
Because the SARS-CoV-2 virus hijacks cells and impact’s the patient’s immune system in similar ways to other coronaviruses, any identified peptide drugs for COVID-19 may also be useful in treating other coronavirus illnesses, like SARS and MERS. A drug to treat lung fibrosis could also be used for other fibrosis-inducing diseases.
“It is obvious that more than one solution may be needed to mitigate the impact of the COVID-19 pandemic and the Nuritas team is eager to leverage our proprietary AI platform,” Khaldi said in the company’s press release.
Originally published at https://www.biospace.com.