In the field of innovative drug research and development, natural products have always been the treasure house of drug discovery - from artemisinin to paclitaxel. With their unique chemical diversity and biological activity, natural products have provided countless possibilities for human disease treatment. However, the drug likeness evaluation and target research of natural products face great challenges for a long time:
Complex structure: natural molecules (such as polysaccharides and alkaloids) often have poor oral absorption or high toxicity due to their high polarity and unstable metabolism.
Data scarcity: more than 80% of natural products lack systematic ADMET (absorption, distribution, metabolism, excretion, toxicity) data.
Fuzzy target: traditional research relies on empirical screening, which is difficult to accurately correlate compounds with disease targets.
Chengdu biopurify Technology Co., Ltd., together with Sichuan Chongqing Artificial Intelligence Research Institute, launched the world's first natural product drug likeness database. Relying on the revolutionary AI platform optadmet, it realized the systematic drug likeness evaluation of natural products + accurate target prediction for the first time.
1、 Core functions of the database: from drug likeness to target, full coverage
1. all AI driven drug likeness prediction (optadmet platform)
Based on Neural Network + multi-dimensional molecular descriptors, covering 160 + key drug likeness parameters, it accurately solves the development pain point of missing drug likeness data of natural products:
| Parameter category | Key predictors | Natural product application scenarios | |
| Absorption (a) | Oral bioavailability (FA), intestinal permeability (caco-2/mdck/peff), PGP substrate propensity, solubility (SW) | Improve the oral absorption efficiency of flavonoids, alkaloids and other molecules | |
| Distribution (d) | Blood brain barrier permeability (BBB), plasma protein binding rate (ppb) | Optimize brain targeting of saponin molecules and reduce non-specific distribution | |
| Metabolism (m) | CYP450 metabolic substrate / inhibition, CYP450 metabolic parameters, CYP450 metabolites, phase II metabolic tendency, metabolite toxicity warning | Blocking metabolic inactivation sites of coumarin molecules | |
| Excretion (E) | Half life (t1/2) | Prolonging the in vivo action time of polysaccharide molecules | |
| Toxicity (T) | Hepatotoxicity (elevated transaminases), cardiotoxicity (hERG inhibition), genotoxicity (Ames test), phototoxicity | Avoiding the risk of liver injury from pyrrolizidine alkaloids |
Technological breakthroughs:
The world's first ADMET parameter for natural products has an accuracy of more than 30% higher than that of traditional tools.
The independent intellectual property model has been validated by FDA approved drug datasets.
2. target prediction and disease association: from molecular to clinical, one-step!
The database integrates five AI prediction tools + real literature data to accurately target the action targets and therapeutic potential of natural products:
Multi tool joint prediction:
Swisstarget prediction (based on 2d/3d similarity)
Targetnet (random forest algorithm)
Npai engine (deep learning target mining)
Open targets (gwas+ drug database validation)
Deepseek (large scale literature mining)
Application scenario:
Known compounds: rapid validation of target and disease association (such as baicalin anti-inflammatory target prediction).
Unknown compounds: AI fills the research gap (such as the prediction of anticancer mechanism of Haiyang natural products).
2、 Five core advantages of database
1. High precision prediction: 160 + drug likeness parameters are fully covered, and the AI model is trained with tens of millions of data, and the result is reliable.
2. Target disease whole chain analysis: from molecular structure to clinical indications, multi tool joint prediction, combined with real literature data, covers the comprehensive information of target and disease.
3. Fill the blank of natural product research: compounds without literature support can also obtain AI prediction data.
4. Cross domain applications:
New drug research and development: rapid screening of high drug likeness natural lead compounds.
Drug repositioning: new use of old drugs and discovery of hidden value of natural molecules.
Functional food: evaluate the safety and efficacy of plant extracts.
3、 Enabling R & D: the "ultimate weapon" for natural product development
Biopurify natural products database can provide the following information for pharmaceutical enterprises, scientific research institutions, and functional food developers:
Lead compound optimization: ADMET parameters guide structural transformation and reduce the clinical failure rate.
Rapid verification of targets: reduce the cost of wet (real) experiments and accelerate project approval.
Modernization of traditional Chinese medicine: use data to solve the problem of "effective but toxic" and promote international declaration.
Interested customers are welcome to contact us. Contact information: Mr. Xie, 13541283596 (wechat), kylin@biopurify.cn .
Examples of database products and parameters:
1. Examples of diseases and targets
2. Drug likeness data example
Drug likeness data include:
TPSA (polar surface area)
LogP (logarithm of oil-water partition coefficient)
Logd (logarithm of oil-water partition coefficient under neutral conditions)
SW (solubility in pure water mg/ml)
Peff (permeation efficiency in small intestine cm/s × 10-4)
Caco-2 (apparent permeability coefficient cm/s × 10-7 on Caco-2 cell model)
BBB (blood brain barrier permeability)
RLM stability (whether the compound is stable in rat liver microsomes)
Hppb (binding fraction of compound in human plasma protein%)
Syn accessibility (availability of synthesis. The higher the value, the more difficult the synthesis is)
MRTD (whether the recommended maximum daily oral dose is higher than 3mg/kg/day)
Rat LD50 (LD50 of half lethal dose of mg/kg rat, the value of drug is generally greater than 300)
Skin sens
Resp sens (does the compound produce respiratory sensitization)
HERG (whether it can inhibit HERG potassium channel)
Chromosomal aberr (whether it causes chromosome variation in mammalian cells)
Photo tox (whether there is phototoxicity) Ames (whether there is Ames toxicity risk. The higher the value, the higher the risk. It is generally believed that Ames toxicity exists when it is greater than 1)
Ser_alk (whether it can cause the increase of alkaline phosphatase level)
Ser_ggt (whether it can cause the increase of γ - glutamyl transpeptidase level)
Ser_ldh (whether it can cause the increase of lactate dehydrogenase level)
Ser_ast (whether it will cause the increase of serum glutamic oxaloacetic transaminase level)
Ser_alt (whether it will cause the increase of serum alanine aminotransferase level)
;