Chemical Drug Development

Drug R&D is a very complicated, costly and time-consuming attempt. But with the advent of AI, this process can be simplified and more quickly paced. AI system allows optimization of an algorithm to identify new chemical matter with a desired molecular profile. AI has been used from the beginning of the drug discovery process, including for initial hits from de novo design generated directly from data.

Chemical Universe

Chemical database (10⁶⁰) is estimated by the molecules containing up to 30 C, N, O and S atoms, and it just a tiny fragment of all the compounds with drug-like properties by logarithmic scale.

Chemical Source for Drug Research

7,895 drugs and experimental drugs
1.7 M bioactive small molecules from ChemBL
~10 M small molecules for HTS
378 M commercially available compounds from ZINC 15 (drug-likeness chemical space)
166 B enumerated small molecules up to 17 atoms (C, N, O, S and halogens)

Chemical Space of Drug-like Small Molecules in Theory

10²⁴ organic molecules with up to 30 atoms with known functional groups from P.Ertl, J,Chem. Inf. Comput. Sci. 2003, 43, 374.
10⁶⁰ Drug-like small molecules up to MW of 500 from R. S. Bohacek, C. McMartin, W. C. Guida, Med. Res. Rev. 1996, 16, 3.
Utilising AI to search a chemical space, which includes potentially billions of options in terms of atom configuration, enabled the researchers to reduce the time taken to identify their target. Integrating data-generating hypotheses with machine learning to produce drug design concepts, the steps previously undertaken by humans were replaced with a suite of advanced algorithms.

AI-Driven Drug Discovery

Compared with traditional methods, CD ComputaBio can help customers save the cost of screening candidates by tens of billions every year. This AI platform can be widely used in various scenarios regarding drug development.

Design novel leads Structure-based Drug Design

Feasibility assessment of new drug targets.
High-throughput screening and active compound discovery based on structural design.
Discovery of active compounds to lead compounds.
Optimization of lead compounds to determination of preclinical drug candidates.
Research on structure-activity relationship.

Fragment-based Drug Design

Feasibility assessment of new drug targets.
High-throughput screening and active compound discovery based on structural design.
Discovery of active compounds to lead compounds.
Optimization of lead compounds to determination of preclinical drug candidates.
Research on structure-activity relationship.

De novo Drug Design

Analyze the active site of the target to determine the distribution of active sites, potential fields and key functional residues.
Use different strategies to put the basic building blocks into the active site and generate completed new molecules.
Calculate the binding energy of new molecules and receptor molecules to predict biological activity.

nomain-title-log-pic1 Screen hits from existing compound library (virtual screening)

Structure-based virtual screening
Ligand-based virtual screening
Reverse Virtual Screening

nomain-title-log-pic1 Molecular Generation and Filtering

Sampling enough examples from the pretrain model (remove duplicates in iterations)
Filtering by -2<logP<6, 350<MW<750, HBD<5, HBA<10, TPSA<150, NRB<10.
Removing molecules which are not pass MCF.
Removing molecules that are similar to those in patent. (Tc>0.85)
Removing molecules that are similar to each other. (Tc>0.85)
Removing molecules that contain scaffold present in the patent.
Adding synthetic accessibility(SA) and Quantitative Estimate of Drug-likeness(QED).
Retaining molecules that are specific in interaction with target.
Manual selection by medicinal chemist.

Timeline

Results and Delivery

Molecules designed by AI exhibit reasonable quality: they are drug-like, synthesizable and within the chemical space of specific target.
The chemical candidates designed by AI exhibit potent bioactivity against specific targets.
The total amount of time for drug design is about 4-6 weeks, including data collection, modeling building, molecular generation and selection.