Unique "Dual-Engine Drive"

Empowered by AI,
Validated by Quantum Chemistry

FMO First-Principles Calculation

Physics-Based Core
 Rejecting empirical parameters, directly calculating Fragment Interaction Energy (IFIE) at the electronic level based on quantum mechanics.
Providing gold-standard validation based on physical laws for screening.

SHAP Explainable Large Model

AI Core
 Introducing game theory methods to quantify and visualize the contribution of every atom to activity.
Moving design from guesswork to rational insight.

AICre Discovery DeepCre Lab
AI Quantum Drug Discovery

End-to-End Data Intelligence Engine

ISO 10993-5 Standard

Standard Curation & Double-Blind Validation

Industrial Data Loop based on Geometric Deep Learning & Active Learning

Multi-modal Knowledge Graph Construction
"Breaking data silos,
building a multi-billion scale pharma knowledge brain."
Multi-modal Knowledge Graph
Multi-source Data Fusion
Core Definition:
Leveraging LLM technology to align global patent, literature, and clinical data at the semantic level, constructing heterogeneous graph neural networks.
Hardcore Tech:
BioBERT + NER, automatically extracting Compound-Target-Pathway relationships.
Data Fusion:
Connecting three major data silos: ChEMBL (Activity), PubChem (Physicochemical), Patents.
Mining Vol.4M+
Precision99.2%
High-Fidelity Chemical Curation
"Garbage in, garbage out?
We only feed AI the purest 'fuel'."
High-Fidelity Chemical Curation
30% Cleaned
70% High-Qual
Core Definition:
Based on QSAR-Ready industrial gold standards, executing a chemical structure cleaning process 10 times stricter than academia.
Hardcore Tech:
Automated cleaning pipeline, including Tautomer standardization, stereochemistry correction, desalting, and neutralization.
Red Line Standard:
Introducing ISO 10993-5 medical device toxicity standards, marking data with cell viability <70% as "absolute negative samples".
Usability99.9%
StandardQSAR-Ready
AI Molecular Representation Learning
"Beyond traditional fingerprints,
letting AI 'understand' molecules like a chemist."
Molecular Representation Learning
1D
SMILES
+
2D
Graph
+
3D
Conf.
Core Definition:
Abandoning traditional 0/1 fingerprints, adopting Graph Neural Networks (GNN) and pre-trained large models to extract high-dimensional features directly from molecular graph structures.
Pre-trained Models:
Introducing MolCLR or ChemBERTa to pre-learn chemical syntax on tens of millions of unlabeled molecules.
Geometric Deep Learning:
Combining 3D conformational information, calculating Geodesic Distance Matrix to capture steric hindrance effects.
Dimension2048-D
Gain+40%
Ensemble Prediction & Active Learning
"Not just accuracy,
but 'enrichment' capability in real-world combat."
Ensemble Modeling & Active Learning
Train
→
Valid
→
OOD Test
Core Definition:
Addressing "small sample" and "imbalanced" challenges using ensemble learning strategies and scaffold split validation.
Scaffold Split:
Forcing the model to predict unseen molecular scaffolds, simulating real-world R&D scenarios.
Uncertainty Quantification:
Outputting "confidence" scores, prioritizing high-confidence molecules for wet lab testing via active learning.
EF1%>30
AUC0.95+
SHAP Explainable Attribution
"Rejecting AI black boxes,
translating prediction results into language pharmaceutical experts understand."
XAI & SAR Decoding
Pharmacophore
H-Bond
Core Definition:
Using Game Theory concepts to generate visualized atomic contribution heatmaps.
SHAP Values:
Calculating the positive/negative contribution of each group (e.g., nitro, halogen) to activity/toxicity.
Value:
Directly guiding optimization, telling chemists to "remove this methyl group" or "keep this aromatic ring", etc.
AttributionAtom-Level
Black BoxTransparent
Dual-Track Strategy

Holistic Drug Design Strategy

For scenarios with known or unknown target structures, we provide adaptive solutions to circumvent limitations of traditional methods.

Get Technical Whitepaper

SBDD Structure-Based Design

Cryo-EMActive Site
  • Atomic-level Precision: Provides atomic-level details of the target protein, facilitating precise drug design.
  • Target Specificity: Designs molecules that specifically interact with the Active Site.
  • Mechanism Insight: Elucidates binding mechanisms and predicts the impact of molecular modifications on binding affinity.

LBDD Ligand-Based Design

SAR AnalysisNo Structure
  • No Structure Dependence: Directly utilizes existing active ligand data when target structure is unknown or hard to crystallize.
  • SAR Analysis: Constructs Structure-Activity Relationships (SAR) to identify key features responsible for biological activity.
  • High Efficiency: Consumes fewer computational resources compared to SBDD, suitable for large-scale rapid screening.

Proprietary Dual Engines: The Loop of Generation & Validation

Internalizing SBDD/LBDD strategies into algorithms, driven by ISO-standard data flows

GenelP 1.3 DESIGN

Generative Intelligent Pharmaceuticals

Designed to solve the pain point of being "limited to known chemical space". GenelP breaks through existing patent barriers to discover Novel Scaffolds unexplored by human experts.

Core Capabilities (Based on Rational Optimization)
Rational Optimization Rational optimization based on target structure to improve efficacy and reduce side effects.
Novel Scaffolds Utilizing generative models to identify novel molecular scaffolds undetectable by traditional methods.
Bias Mitigation Overcoming LBDD's bias towards known chemotypes, expanding chemical space.

ChscIP 1.2 VALIDATION

Chemistry Screening Intelligent Platforms

A high-precision prediction engine built on rigorous data flows. ChscIP combines ISO Standards with Explainable AI, solving the "black box" and "poor generalizability" problems of traditional prediction.

Data & Validation Logic (Source: Real Data Flow)
ISO 10993-5 Standard Strictly following ISO standards, defining cell viability < 70% as the toxicity red line.
Feature Engineering Constructing 4 continuous + 32 categorical features, transformed into Dummy Features for model input.
Proof of Generalizability Adopting Nested Cross-Validation and independent External Validation for double insurance.
SHAP Values Outputting SHAP values to identify Key Attributes, achieving explainable prediction.
Standard Workflow

AI-Empowered · Standardized Drug R&D SOP

From Target Discovery to Preclinical Candidate (PCC)

01
Target Identification & Confirmation

Target Identification

AI Analysis: Omics Bioinformatics Analysis
Wet Lab: Multi-omics Detection
02
Druggability Assessment

Druggability Assessment

AI Prediction: Pocket & Allosteric Site Prediction
Wet Lab: SPR/BLI Affinity Screening
03
Structure Modeling & Preparation

Structure Preparation

AI Modeling: AlphaFold2/3 Complex Prediction
Wet Lab: Cryo-EM & Protein Crystallization
04

Hit Discovery

Hit Discovery
AI Core Intervention: Generation & Screening Dual Engines
Path A: De Novo Design

Breaking patent barriers, creating brand new scaffolds

Generative Models (SBDD) Scaffold Hopping FBDD Fragment Assembly
AI: Generating 10^6 scale high-potential library
AI: SA Score (Synthetic Accessibility) filtering
Path B: Ultra-Large Scale Virtual Screening

Mining existing active molecules from billion-scale libraries

DEL Library Screening Pharmacophore Matching Ultra-fast Docking
AI: Billion+ library processing in seconds
AI: 3D-CNN Conformational Scoring
05
Lead Optimization

Lead Optimization (Hit-to-Lead)

How AI Accelerates Optimization?
FEP Free Energy Perturbation:
 Achieving chemical accuracy (1 kcal/mol) in affinity calculation, replacing expensive synthesis.
Multi-Parameter Optimization (MPO):
 Dynamically balancing Activity / Solubility / Toxicity / Metabolic Stability.
ADMET Assays: Liver Microsome Metabolism, Membrane Permeability (PAMPA)
Safety Evaluation: hERG Cardiotoxicity, Ames Genotoxicity
06
Wet Lab Validation & PCC Selection

In Vitro / In Vivo Validation

DeepCre Lab Delivery:
We provide full-chain data delivery from synthesis to animal models, ensuring data traceability.
Animal Efficacy: CDX/PDX Mouse Models, Tumor Inhibition Rate
PK/PD: Pharmacokinetic Parameters (Half-life, Bioavailability)

Proprietary Pipeline

High-potential assets incubated by ChscIP platform · Seeking global partnerships

AICRE-U01

Gout / Hyperuricemia URAT1 Inhibitor
Current Stage: Non-clinical Animal Studies
Target ID
Hit ID
Lead Opt
PCC
IND
2,638 Entries
High-Quality Training Data
Patent & Lit. Curation
0.94 AUC
Model Accuracy
LightGBM Ensemble
330 M+
AI Screened Library
ZINC15 Lead-like
4 Sites
Key Allosteric Sites
Structural Mechanism
AI-Empowered Breakthroughs
  • Rapid Screening Funnel:

    Using ChscIP, rapidly narrowing down 3.3 million molecules to 411 high-potential candidates (SHAP >= 2.35) via AI scoring and docking, increasing efficiency by 500x.

  • Explainable Attribution:

    Identifying core pharmacophores of Top 8.3GB via SHAP analysis, precisely eliminating ineffective scaffolds.

Mechanism Elucidation & Differentiation
  • Novel Mechanism Discovery:

    Revealing dynamic conformations of the URAT1 transport tunnel via MD simulations, discovering gating regions TMD7 (W357) and TMD11 (R487).

  • Atomic Differentiation:

    Pinpointing the positive charge effect of residue R477, providing atomic-level basis for designing high-selectivity, low-side-effect inhibitors.

AI+ Drug DISCOVERY

AICre Discovery Engine

Reshaping drug discovery with algorithms · Full-stack computational services from hit screening to property prediction

High-Throughput Virtual Screening

Based on ChscIP platform,
performing SBDD/LBDD bidirectional precision mining on billion-scale libraries.

  • HTVS: Billion-scale library docking in seconds (ZINC/Enamine)
  • Hit ID: Hit identification & Scaffold clustering
  • Repurposing: Drug repurposing & Indication expansion

Generative Molecular Design

Utilizing GenelP generative models to break patent barriers and expand novel chemical space.

  • De Novo: De Novo design for target pockets
  • Scaffold Hopping: Scaffold hopping for patent circumvention
  • PROTAC: Smart design of Linker length & conformation

Property Prediction & Optimization

Fusing Quantum Chemistry (QM) with Deep Learning,
providing prediction accuracy comparable to wet labs.

  • FEP Calculation: FEP high-precision affinity prediction
  • ADMET: Metabolic stability & BBB permeability
  • Toxicity: hERG cardiotoxicity & Ames prediction

POWERED BY: PyTorch AlphaFold3 Gromacs RDKit DiffDock

DeepCre Experiment Center

AlCrepharm's physical operation center in Sino-Japan (Tianjin) Health Industry Park
A research hub connecting global resources

STRATEGIC PARTNER Fujifilm & JCRB Cell Bank

Wet Lab Validation

In Vitro Validation
Promise:
All experiments use authentic JCRB official cells
  • Chemical Synthesis
  • ISO 10993-5 Cytotoxicity Test
  • IC50/EC50 Efficacy Curve Determination
  • Western Blot / qPCR Mechanism Validation
  • Reporter Assay

Translational Research Services

Research Translation

Relying on the young scientist team from Osaka University,
providing high-level translational support for universities and hospitals.

  • Innovative project design & Logic structuring
  • Delivery of SCI-publication-grade raw data
  • Patent layout & Technology transfer guidance
  • Pharma pipeline docking (License-out)

Strategic Resource Import/Export

Global Sourcing
FUJIFILM Deep Cooperation
JCRB Cell Bank Deep Cooperation
  • Compliant cold-chain import of biologics/reagents
  • Exclusive sourcing of scarce Japanese cell lines
  • Global procurement & Import/export agency
Contact DeepCre for Resource Catalog
Osaka University Suita Campus
Young Scientists Elite PhD Team
Originating from Osaka University · Young Scientist Pioneers

Intelligent Synthesis, Creating the Future of Medicine
A Pioneer Team of New Generation Young Scientists from Japan

AlCrepharm originated from Osaka University.
Our team includes several outstanding young scientists from Japan. As an AI+ tech enterprise, our core team members possess dual degrees in AI and Pharmacy.

Core Philosophy:
Rejecting blind "big data alchemy", we adhere to a "First Principles" drive.
Introducing "Fugaku" supercomputer-level FMO calculation methods to analyze life phenomena at the electron cloud level.
Dedicated to building a "Dry-Wet Loop" paradigm for next-generation drug R&D.

92.3%

Structure Prediction Accuracy

90%

Team Master/PhD Rate

1/3

Shortened R&D Cycle

"Empowering Drug Discovery with Quantum Mechanics & Youthful Innovation."

Company News

Latest R&D progress and industry milestones of AlCrepharm

LATEST UPDATES
GenelP 1.3
R&D Update
2025-09-01 · AICre Tech
GenelP 1.3 Major Release

New version introduces equivariant diffusion models, improving scaffold novelty by 40% and significantly optimizing SA scores in molecular generation tasks.

JCRB Partnership
Partnership
2025-08-15 · DeepCre Lab
Strategic Partnership with Fujifilm

Strategic partnership reached. All DeepCre wet lab experiments will use authentic, traceable JCRB cells.

URAT1 Progress
Milestone
2025-07-20 · Drug Discovery
AICRE-U01
Gout Project PCC Selected

Based on 22 hit molecules locked by ChscIP platform,
wet lab validation shows Top 10 molecules exhibit excellent URAT1 inhibitory activity.