Computer modeling of new drugs
Two times faster, six times cheaper: modeling drug candidates before the lab
A screening pipeline that starts with a protein structure and ends with a ranked list of compounds worth synthesising: Hanna Karpenka, team leader of the Drug Design Group at Belarus's United Institute of Informatics Problems, says it runs at least two times faster and six times cheaper than classical discovery. Her group has already flagged candidate inhibitors of the SARS-CoV-2 main protease. Where exactly do neural networks enter the chain?
See the five-step pipeline that turns a protein target into testable drug candidates
In short
Hanna Karpenka presents how her computational drug design team turns a known biological target into a ranked list of compounds worth synthesising, using pharmacophore models, virtual screening, molecular dynamics and neural networks. She walks through the group's five-step workflow, its published COVID-19 results and the kinds of research collaboration it is looking for. Her argument is that a theoretical, computer-first approach cuts both the time and the cost of early drug discovery.
“As you can see, I and my team are professional from various benches of science such as chemistry, computer science, biology, mathematics and physics.”
“On the next step, we use obtain results as a teaching dataset for our neural network and generate more potential active compounds.”
What you will learn
- Follow a five-step computational workflow from biological target to candidate compound list
- Spot where neural networks add value in early-stage discovery and where classical methods still decide
- Compare the time and cost of computer-aided screening with laboratory-first discovery
- Understand what a known, deposited protein structure makes possible for a modeling team
- Plan a research collaboration around joint projects, publications and patents
Key takeaways from the speaker
- bioinformatics
- drug design
- ML
Details
- Level
- Intermediate
- Format
- Lightning talk
- Technical depth
- Introductory technical
- For
- Early career, Mid-career, Senior individual contributor
- Length
- 5:48
- Published
- June 2021
Questions this talk answers
What are the steps in the team's computer-aided drug discovery workflow?
A target is chosen, a pharmacophore model is reconstructed and virtual screening is run, identified compounds are sorted with molecular modeling tools, and those results become a training dataset for a neural network that generates more potential active compounds. The best compounds are then evaluated with molecular dynamics and quantum chemistry tools, which the speaker says give the most reliable results.
Where do neural networks fit into the process today?
The speaker says neural networks can in principle be used at every step, and distinguishes on her diagram the green part the team already does from the blue part it expects to add.
How much time and money does the computational approach save?
Because the approach is theoretical, it is less cost and time demanding: at least two times faster and six times cheaper than the classical approach, as the speaker states.
What does a team need before it can model a disease target?
The group can work on any disease whose biological trigger is known and deposited in the Protein Data Bank.
What is the end product of the work?
A list of potential inhibitors for a certain biological target, with detailed information about their activity, ready for laboratory testing in vitro and in vivo.
About this talk
omputer molecular modeling techniques play an important role in research into the development of new drugs. The use of neural networks for these methods in the rational design of potential drugs can minimize the time and cost of searching for various targets and potential drugs! Such a computer model can, with high probability, predict the vulnerability in the structure of the infectious agent and indicate which chemical compound should be synthesized.
Topics
Highlights
- From target to candidate compounds in five steps (60 s)
- Faster and cheaper than classical discovery (58 s)
- The market pushing drug discovery to scale (49 s)
Good afternoon, everyone. My name is Hanna Karpenka and today I would like to present to you my project: Computer discovery of novel drugs. The technologies used as a project are progressive and used worldwide. Recent research shows the drug design market is activity development. The global drug discovery service market size is projected to reach 21 billion dollars. By 2025, by 11 and 1 billion dollars in 2020. And CIGR. Of 14% during the forecast period. Growing everyday expenditure in the pharmaceutical and bio-pharmaceutical industry. This necessitates an increasing demand of outsourcing, analyzing, testing and clinical trial diseases.
initiatives for research on rare diseases or often drugs and focus on the drug discovery are driving the growth of the global drug discovery market through the industry. Growing in biology, patent experience and imaging economics is expected to provide a wide range of growth opportunity to players in the market. As you can see, I and my team are professional from various benches of science such as chemistry, computer science, biology, mathematics and physics. That is why we are capable of working with multi-disciplinary projects such as computer-added drug design.
Our team has extensive experience in the using such molecular modeling tools. As molecular docking, molecular chemical research and so on. In recent years, our team has identified series of novel HIV anti-inhibitor scapholds, the number of kinase-inhibitors and so on. However, recently we have discovered small drugs compounds as potential inhibitors of SARS-CoV-2 main projects. What is the more important, we developed a future-aware approach in order to identify potential-efficient inhibitors to new biological targets.
We already had successful projects with the Fudan University, joint European initiatives and the Armenian Academy of Science. Our key publications were published in high-certated international journals. The first results of our work on COVID-19 issue were published in the journal of bioinformatics and computing biology with impact factor higher than 3. This is the present day the article issued about more than 2,000 times and downloaded more than 600 times.
We consider presentation and population of our result is extremely important part of our work. Our team members actively participated in the conference and scientific competition. This drives the maintain of our competence of international level. As a result, our team is among the finalists of the International JEDRI COVID-19 Challenge and the final I expected in August. So, here I would like to tell you about our main work scheme.
On the first step, we choose a target that reconstructs the pharmacopharm model and carry out virtual screening on the basis of it. Then we sort identified compounds in order then efficiently by the means of molecular networking tools. On the next step, we use obtain results as a teaching dataset for our neural network and generate more potential active compounds.
The last step in the evaluation of the best compounds with molecular dynamics as well as quantum chemistry tools since its method gives the most reliable results. Additionally, we provide a comprehensive description of binding profilis and undermed properties for the best compounds. In a nutshell, the result of our work is a list of potential incubators to this certain biological target that are to be synthesized and tested extremely accompanied with the detailed information about their activity.
I have to tell you that nowadays we can use other neural network on all steps which you can see from our area. The green part is what we are doing, the blue part is what we expect. Since our approach is a theoretical one, is at much less cost and time-demanding. It's at least two times faster and six times cheaper in comprehension to the classical approach.
As you can see from the table of the slide, every step of lead structure discovery is significantly accelerated by the means of molecular monitoring and chemoinformatics. The type of cooperation that we are looking for is a joint research project, joint publication in the international journal and joint patent. So, what we can offer? We are offering a full cycle of computer drug discovery cycle as well as some particular expertise in any of the film methods above.
Of any diseases for its biological trigger is known and deposited in the protein data bank. To sum up, the result of our work is the lead of potential inhibitors to certain biological targets. This is the drug candidates are ready for laboratory expertise for the future in vitro and in vivo modification. So, thank you for your attention. We'll be glad to listen and to get your questions here or by email, link it in or if you would like to visit our website.
Transcribed automatically; names checked against the speaker record. Select a time to jump there.
Frameworks and ideas mentioned
- Pharmacophore modeling
- Virtual screening
- Molecular docking
- Molecular dynamics
- Quantum chemistry methods
- Chemoinformatics
Resources mentioned
- Protein Data Bank
- Journal of Bioinformatics and Computational Biology
Cite this talk
HANNA KARPENKA. "Computer modeling of new drugs." WomenTech Network, 2021. https://www.womentech.net/video/computer-modeling-new-drugs