Transcript DrugDiscovery@Home - Digital Bio Pharm

DrugDiscovery@Home
The presentation is prepared by:
Andrey Voronkov, PhD – speaker (MIPT, Lomonosov Moscow
State University)
Vladimir Barinov – Grid Dynamics
IDEA OF THE PROJECT
Computational chemistry + Computational biology + Open code +
People’s masses (volunteer computing) = New level of pharmaceutics and biotechnologies
Filter for toxicity - filter for huge databases including virtual
chemical spaces QSAR, side effects
Filter for pharmacokinetics
Bioinformatics and OMICS data - selection of biotargets
Docking and virtual screening. Enrichment (filtration
according to activity)
Molecular dynamics, FEP, TI
Retrosynthetic analysis, modeling of synthetic availability.
Result – a lot of compounds which fast pass preclinical and clinical trials
Possibility to create multitarget drugs.
«The problem» of the consumer. What problems of
the target group does the project solve?
1)
Problems of the consumer (pharmaceutical companies, drugs development:
-
High prices (above 1 billion of dollars) and long period (10-15 years) are required
for the development of new medicines
There is a need in new medicines
Impossibility to study the virtual chemical areas (compounds which may be a
drug but are not synthesized yet – these is no less that several hundreds of billions
of such compounds)
2)
-
Current solutions of the problem:
Biological experiments
Use of leased cluster
Сloud computing
3) Shortcomings: expensive and long. Lack of the community involvement and
crowdsourcing.
Target groups:
Large, medium-sized and small pharmaceutical companies
Academical research organization leading preclinical development of drugs
Product and technology– BOINC, @HOME
The method allowing to solve the problems: experiments are
simulating in virtual area.
Additional description of the technology:
- It is the fact that resouces of modern computers /
tablet computers / smartphones are in sleep mode
and not used for the essential part of time.
- Volunteers give their computer resources (various
motivations) and we obtain scalable highperformance network.
- Segmentation of tasks into subtasks. Validation.
Description of the product and project technology
Product is the platform for distributed
computing that uses volunteer computer
resources for new drugs development.
Client
Project parts
Software:
- Software with open code for drugs
modelling – from the identification of
biological targets and computer models of
proteins to the pharmacokinetics.
- At the moment: software for doking
(modelling of the interaction between
chemicals and proteins) and molecular
dynamics.
- Then additionally, we are planning to
install software for the proteomics,
metabolomics and sequencing data
analysis.
Server
BOINC is the platform with open
code that provides interaction
between client and server parts
of the project.
Shortcomings and criticism
These methods went out of date and they are inferior to experimental methods of high-throughput screening,
or DNA-tag etc.
These methods did not go out of date. Statistics shows that the popularity of these methods grows up.
However, today this process is not so fast as at the beginning of their use, or it is not so fast as in some other
areas, e.g. in biostatistics. But our project also implies that analysis of huge OMICS data will be used for
biotarget identification. Our project should be the most price-competitive comparing to any rivals in the area
of preclinical trials.
Methods are inefficient. In this case, it is most often referred to docking.
Docking is effective not for the selection of prospective compounds, but for the sample enrichment, actually,
it is one of the filters. At the output of this filter we will set more effective and resource-intensive methods,
such as molecular dynamics, perturbation of free energy and thermodynamic integration.
Volunteers could refuse to count a commercial project.
Our experience shows that it is not true. We have hired thousands of computers in the first weeks of alpha
testing.
Cloud computing or clusters could replace volunteer computing.
They could not. Volunteer computing will always be cheaper.
Users replace notebooks and personal computers with smartphones.
Smartphones get more and more powerful, and they keep their battery charge longer and longer. There is a
possibility to conduct volunteer computing on smartphones.
It is difficult to keep confidentiality while we use volunteer computing. Confidentiality is important for
aggregate results only, and these results will be protected on the server. Confidentiality is unimportant for
intermediate results, but it likely still be kept due to huge volume of data.
Currency of the project’s technologies
QSAR
Moreover, the great part of publications in a number of core journals, such as J Med Chem, ChemMedChem
include certain methods of computer modeling of drugs.
By now, there is a lot of drugs which were developed using computer modeling methods.
Talele T.T., Khedkar S.A., Rigby A.C., Successful applications of computer aided drug discovery: moving
drugs from concept to the clinic. // Curr Top Med Chem. 2010;10(1):127-41.
Virtual screening using molecular dynamics
About 12 hours for GPU, for 1 compound. We are planning to get involved 20-30 thousands of
GPU. It provides the same speed as in high-throughput screening.
PC – protein with active ligand, NC – protein without ligand, o212 – ligand attitude
Business model
How does the project earn and plan earnign?
1. Creating new medicines (patents and royalty). Preclinical trials
as fast as possible.
2. Working by contract as Contract Research Organization (CRO).
3. Leasing of the computer resources.
4. Solving problems by request using crowd-sourcing (getting users of
the platform involved).
Distribution (the way to the final buyer):
- Direct B2B sales
Life hacking:
- Crowd-sourcing for the portion of work in the areas of project
development and forwarding.
- Search for polypharmacological (multitarget) drugs, a new, more
promising paradigm. It is more difficult to use high-throughput
screening for multitarget drugs creation.
Consumer and market
- The final market of the project is the pharmaceutical R&D
outsourcing market. The project will issue it with preclinical products
(amount of money is 27 billions of dollars in 2014):
http://www.contractpharma.com/issues/2014-1101/view_features/adoption-of-fte-contracts-in-rd-outsourcing-presentand-future).
The perfect desirable scenario would be like that:
- Several hundred thousands of dollars in two years
- Several billions of dollars in 5-6 years
- Steady increase up to occupying a leading market position
Market share you are planning to occupy – this is difficult to forecast
(perfectly - 100 % :-)). A minimum plan is any possible market share)).
How far is the market competitive?
- The market is highly competitive with a pent-up demand though.
Business rivals. Who else is solving the same or an
adjacent problem?
Technologically, the closest projects are Docking@home и GPUGRID, which are nonprofit
and not engaged in new drugs creation.
1. The Docking@home Project performs docking of low-molecular compounds. It is
closed. The docking proper is ща little avail. World Community Grid – mainly, docking.
2. The GPUGRID Project uses GPU processors to model proteins and mechanisms of
protein–ligand interactions. It is used for high-throughput molecular dynamics. It has
good customer reviews. It has standing orders.
3. Several hundreds of companies and research groups are working in the area of
preclinical development of drugs in various countries.
What are your advantages?
The main distinction and advantage of our project is the complex approach to the
development of drugs and presence of a great number of applications that allow to
solve problems of computer design of drugs, molecular modeling and docking. We
also aim at the analysis of genome sequences, usage of collective intelligence and crowdsourcing for solving the broad range of the problems. Intrinsically, the project proposes to
get great number of users involved into the process of drug creation.
Traction
METRIC 1 The amount of users and expected resources.
METRIC 2 Contract volume of R&D services (this includes
obtaining grants and financing for the project from the state and
from government contracts).
METRIC 3 The amount of patents that are filled up and research
papers published
Current characteristics of the project:
Current users: 1769
Curent computers: 5748
Development policy
Key points in the project’s development (schedule plan)
- Creating a workflow
- Advanced: testing and validation of the methods,
workflow in experiments for testing drugs
- CRO – rendering a service in drug development and
computer resources providing
- Development of drugs by ourselves
Team
Voronkov Andrey, PhD in Chemistry, preclinical development of
drugs
Zaslavskiy Mikhail, PhD, bioinformatician, mathematician,
specialist in machine learning, one of the winners,
AstraZeneca Dream Challenge
Barinov Vladimir, programmer, DevOps
Structure of expenses (investment required)
-
Salaries (for server administrator and programmer, dev-ops is first of all)
-
Server
-
Marketing (it may turn out unnecessary – in case of viral dissemination of information)
Thanks for your attention!
Distributed computing project
www.drugdiscoveryathome.com
Andrey Voronkov, PhD, R&D, drug design and development
[email protected], Tel: +47 46 22 77 96; Skype:digitalbiopharmcom
Mikhail Zaslavskiy, IT, server administration, programming, bioinformatics, machine learning
http://www.zaslavskiy.org/aboutme.htm
Vladimir Barinov, IT, server administration, programming,
[email protected], Mob: +7-903-380-68-30