Why Conventional AI Drug Design Saves Little Cost and Time
The dream of elegantly obtaining the best potency by a small number of design does not come true.
Costly, Inefficient Drug Design
Example workflow of a drug design team:
- Synthesize up to 20 compounds per month
- Synthesis (S) plus bioassays (B) cost €500,000 for 200 compounds
- It takes 200 steps to bring IC50 from 0.2 mM to1.0 µM
- Optimization to 0.1 µM takes：
- 2000 design, S, and B
- 5 million euro
- 8.3 years of development
Limitation of Pipeline Co-Development
Typical workflow of an AI-first partner:
- Spend 3-6 months on AI-driven compound design
- Nominate 200 compounds for S and B
- Repeat that until contract due date and see if the milestone potency is achieved
When this procedure is taken, the time and cost of S and B remain unchanged.
Inflexibility of AI-First Service Model
Requires disclosure of entire information about the pipelines for "co-development", sharing of intellectual property ownership, and royalty payment.
How our Predictive Intelligence Boosts 500x Productivity
A subscription to maximize the number of design at the minimum cost steers you to the desired potency.
200 Bioactivity Yields per Month
- After subscription, we receive 200 structures, predict bioactivities and report to the client
- Client chooses 4 compounds for S and B
- Given the 4 wet-lab results, we re-tune the client-specific model and predict calibrated bioactivities of the remaining structures
- Client obtains 200 bioactivities in a month.
Escalated Productivity
- 1 month of subscription gives 200 bioactivity yields at the cost of 4 tests
- 10 months give 2000 yields at the cost of 40 tests (€100,000)
- 4 years of subscription brings：
- 2 leads
- 1 preclinical candidate
- 1 IND-enabled
- 1 drug in Phase 1
Synergy of 50x Cost Cutting and 10x Time Saving*
- Research cost from 5 to 0.1 million
- Development time from 8.3 years to 10 months
- 4-year deliverables: zero → 5 pipelines
*See effective savings upon the reduction of development time
Maximal Ownership, Minimal Disclosure
Lean* prediction frees you from granting property co-ownership, royalty obligation, and excessive information disclosure.
See how our technology makes it possible.
*Please be informed that we do not provide any service involving the target, drug discovery, molecular design, and synthesis
Plans for Boosted Drug Design
Basic
Standard
Prediction based on the foundation model published in Nature Machine Intelligence
Basic
Advanced
Try our latest model— built on a significantly larger dataset, and validated through a more realistic benchmark system
Basic
Premium
FAQ: We can use your open-source code to do the prediction on our own. What is the advantage of subscribing to your Standard Plan?
You can access the published version of our foundation model in the code but have to fine-tune the model for your pipelines.
The expertise it takes to have fine-tuning work right goes beyond a collection of deep learning algorithms, data sciences, and pharmacology. The validated strength to put theories into effect is available in our service.