Precision.AI

AI Platform: Precision.AI and DrugTX.AI key components

One AI platform with two engines: Precision.AI for target and drug discovery, and DrugTX.AI for development, manufacturing and commercial operations. Each works on its own or as part of one connected suite.

Target and drug discovery

Discovery and clinical analysis

AI-enhanced target generation using publicly available datasets, proprietary datasets from patients and partner companies, and data from genomic alterations linked with responses.
Clinical analysis: adverse event monitoring, endpoint measurement and trial monitoring.
QSAR models to analyze correlations between chemical structures and biological activities for lead selection and optimization.
Molecular docking simulations to predict binding interactions between small molecules and target proteins.

AI-driven antibody and ADC engineering: oncology and CNS differentiation

Molecular dynamics, quantum chemistry and AI-driven mutagenesis integrated to accelerate next-generation antibody and ADC development.
Multi-scale modeling: atomistic and coarse-grained MD to map binding and conformational flexibility.
AI-powered mutagenesis: predictive optimization of CDR loops and linker-payload stability.
Free energy ranking (FEP, MM-PBSA/GBSA): quantitative affinity and stability prediction.
High-throughput variant design: AI-guided ranking before wet-lab synthesis.

How Precision.AI works

DrugTX.AI

Development, manufacturing and commercial

Intelligent asset search and prioritization

Scans 14,000 available assets at various phases of development, such as Phase II and Phase III, and performs deep research analysis.

Software as a service and training

Data Management System (SaaS)
Shipping and Inventory Management Systems (SaaS)
AI-Based Demand Forecasting System (SaaS)
VR-based trainings for biotechnology, nursing and manufacturing

Clinical AI and services

AI-based bronchopulmonary dysplasia (BPD) prediction
Medical insights automation
AI-based tumor and imaging analysis
Professional services

Explore DrugTX.AI

From the target to the preclinical candidate

Each stage combines data-driven models, physics-based modeling and expert review, so weak candidates are filtered out early and lab resources go to the most promising options.

01

Target discovery

Multi-omics, patient and public data, scored by proprietary algorithms for tumor specificity, safety and druggability.

02

Small molecules

Over 6 billion compounds screened with machine learning, docking and molecular dynamics.

03

Antibodies and ADCs

Physics-based multiscale modeling and AI-driven mutagenesis to optimize binding and stability.

04

Lab validation

Focused in vitro validation with scientific experts before any candidate advances.

Lab results feed back into the models in a closed loop, improving predictive accuracy.

01

Target discovery and validation

A proprietary AI-powered bioinformatics platform that identifies novel cancer targets, with the specificity, safety and efficacy that better cancer therapies need.

What goes in

RNA-seq gene expression and proteomics from cancer and cancer-free samples
Public resources such as TCGA and GTEx
Exclusive datasets from partner hospitals and academic collaborators
Data from genomic alterations linked with response to treatment

How each target is scored

Tumor versus normal expression
Surface localization
Essentiality for tumor survival
Internalization behavior
Prognostic relevance
Low expression in healthy tissues

Confirmed before it counts

Every candidate target is benchmarked against current competitors using clinical trial databases and patent landscapes, then confirmed with proteomic and biochemical studies in human samples. Single-cell and spatial technologies check that a target is truly tumor-specific, avoiding immune or stromal cells, a critical step to prevent toxicity.

02

AI-guided small-molecule discovery

Machine learning, deep learning, expert knowledge and high-performance computing across the early discovery pipeline, particularly for difficult targets in oncology.

Lead identification

01
Compound library screening Over 6 billion compounds are screened using machine learning algorithms.
02
Compound filtering Internal tools remove patented or unsuitable molecules.
03
Compound formatting Remaining candidates are converted into SMILES format for modeling.
04
Protein structure processing Binding sites are defined through expert-guided analysis.
05
Docking assay AI models perform molecular docking to predict interaction strength.
06
Post-processing Docking poses are refined with convolutional neural networks and deep learning.
07
Molecular dynamics Simulations assess the stability and feasibility of each interaction.

Lead optimization

AI-based modification

Machine learning algorithms suggest chemical changes to enhance efficacy and safety.

Consecutive docking

Modified compounds are validated using three different docking software tools.

Final in vitro validation

Optimized leads are retested in the lab to confirm improved biological activity.

Methods across the pipeline

Deep QSAR models

Correlate chemical structure with biological activity, including graph-based models that read molecules as atoms and bonds.

De novo design

Generative models (GANs, VAEs) and reinforcement learning propose novel molecular structures beyond existing libraries.

ADMET and toxicity prediction

Flags compounds likely to cause toxicity or adverse effects, so unfavorable candidates are stopped earlier.

Structure-based design

AlphaFold-predicted protein structures reveal binding sites, including on targets once considered undruggable.

03

AI-driven antibody and ADC engineering

A physics-based multiscale platform for antibody optimization and de novo design, focused on oncology and CNS differentiation.

Physics and AI combined

Molecular dynamics, quantum chemistry and AI-driven mutagenesis, integrated to accelerate next-generation antibody and ADC development.

Multi-scale modeling

Atomistic and coarse-grained molecular dynamics map binding and conformational flexibility.

AI-powered mutagenesis

Predictive optimization of CDR loops and linker-payload stability.

Free energy ranking

FEP and MM-PBSA/GBSA give quantitative affinity and stability predictions.

High-throughput variant design

AI-guided ranking of antibody variants before wet-lab synthesis.

Built for internalization

Antibody–antigen interactions, binding energetics and the behaviors that drive internalization, a key requirement for ADCs, are modeled before experimental screening.

04

Validated in the lab, with experts in the loop

Once promising compounds are identified computationally, they are validated in the laboratory with scientific experts, to assess biological activity, efficacy and safety in controlled in vitro settings. Only the most viable candidates advance to the next stage.

Computational hypotheses are refined through focused in vitro validation, so ineffective candidates, both small molecules and biologics, are filtered out earlier and experimental resources go to the most promising options. Lab results feed back into the models in a closed loop, continuously improving predictive accuracy.

The platform is designed as a human-in-the-loop decision-support system, where computational insights inform experimental design and optimization cycles rather than replacing biological validation.

Therapeutic focus

Precision.AI is applied to disease modeling and target discovery in four areas.

Oncology

Pancreatic cancer and non-small cell lung cancer

Children's diseases

Pediatric lung disease

Alzheimer's disease

Neurology

Specialized immunology

Systemic lupus erythematosus

Responsible AI in drug discovery

Precision.AI is aligned with the risk-based credibility framework in the FDA's January 2025 draft guidance, “Considerations for the Use of Artificial Intelligence to Support Regulatory Decision-Making for Drug and Biological Products.”

Defined context of use

A modular design supports a credibility assessment for each model's specific context of use.

Fit-for-use data

Relevant, reliable and representative datasets, with bias mitigation in training and evaluation.

Explainable outputs

Explainability tools such as SHAP values show what drives each prediction.

Monitored over time

Ongoing monitoring for data drift, with lab results feeding back into the models.