About
Apheris is a specialized AI infrastructure platform designed for pharmaceutical and biotechnology drug discovery teams. It powers the industry's largest federated data networks, enabling multiple organizations to collectively train and improve AI models without ever exposing their proprietary datasets to one another. The platform supports three core engagement modes: deploying and running models locally within a secure environment, joining existing industry-wide federated networks, or building custom federated collaborations across partners and sites. Teams can benchmark model versions against curated internal datasets, fine-tune models for specific targets and chemotypes, and integrate outputs via a GUI or API. Apheris powers several major federated networks, including the AI Structural Biology (AISB) Network—a collaboration among AbbVie, AstraZeneca, BMS, Genentech, Johnson & Johnson, Sanofi, Takeda, and others—as well as the ADMET Network for drug metabolism and pharmacokinetics and the Antibody Developability Network. Its flagship product, ApherisFold, enables local deployment of co-folding models with fine-tuning on proprietary structural biology data. Apheris is purpose-built for drug discovery workflows, addressing the need for model reliability transparency, in-house data compatibility, and adaptability as programs evolve—all within a certifiably secure and compliant architecture. It is ideal for computational biology teams, data scientists, and research leads at mid-to-large biopharma organizations.
Key Features
- Secure Local Inference: Deploy and run AI models entirely within your own environment so that data, queries, and outputs never leave your organization.
- In-House Benchmarking: Compare model versions against curated internal datasets to assess reliability and suitability for your specific targets before deploying.
- Model Fine-Tuning & Customization: Fine-tune models on proprietary data for specific targets and chemotypes, then integrate results via GUI or API into existing discovery workflows.
- Federated Data Networks: Participate in or build industry-wide networks (AISB, ADMET, Antibody Developability) that train superior AI models across multiple pharma organizations without sharing raw data.
- ApherisFold: A co-folding model application for protein-complex prediction that can be used independently or extended through federated networks for enhanced accuracy.
Use Cases
- Pharmaceutical teams running AI co-folding models locally for protein-complex structure prediction without exposing proprietary structural biology data.
- Drug discovery organizations joining the ADMET Network to design more informative compound batches and optimize experimental capacity in DMTA cycles.
- Biopharma companies participating in the AISB Network to access and contribute to state-of-the-art AI models trained across industry-wide datasets.
- Antibody R&D teams leveraging the Antibody Developability Network to advance developability assessments using purpose-built federated AI models.
- Computational chemistry groups fine-tuning pre-trained models on in-house proprietary data for specific drug targets and chemotypes, integrated into existing discovery pipelines via API.
Pros
- Privacy-Preserving Collaboration: Federated learning architecture allows multiple biopharma organizations to jointly train superior models without ever exposing proprietary datasets to one another.
- Enterprise Pharma Network: Access to a pre-built network of top-tier pharmaceutical partners including AbbVie, AstraZeneca, BMS, Genentech, Sanofi, and Takeda for unmatched data diversity.
- Purpose-Built for Drug Discovery: Designed specifically for the reality of drug discovery workflows — including model reliability assessment, in-house data benchmarking, and program-level adaptability.
- Certifiably Secure Infrastructure: Comes with a Trust Center and security certifications, making it suitable for highly regulated pharmaceutical environments.
Cons
- Highly Specialized Use Case: Apheris is exclusively focused on pharmaceutical drug discovery, making it unsuitable for general AI model deployment or other industry verticals.
- Enterprise Pricing: Targeted at large biopharma organizations, the platform likely carries significant enterprise-level costs with no publicly available free tier or self-serve pricing.
- Requires Organizational Buy-In: Joining or building federated networks requires agreements and participation from multiple stakeholders, which can slow onboarding for smaller teams.
Frequently Asked Questions
What is federated learning and how does Apheris use it?
Federated learning is a technique where AI models are trained across multiple data sources without centralizing the data. Apheris uses this to allow pharmaceutical companies to collaboratively train better models on combined industry data while keeping each organization's proprietary datasets local and private.
What is ApherisFold?
ApherisFold is Apheris's co-folding model application for protein-complex prediction. It can be used independently of any federated network, allowing teams to run models locally, benchmark on in-house data, and fine-tune on proprietary structural biology data.
Which pharmaceutical companies are part of the AISB Network?
The AI Structural Biology (AISB) Network includes AbbVie, Astex, AstraZeneca, Boehringer Ingelheim, Bristol Myers Squibb, Genentech, Johnson & Johnson, Sanofi, and Takeda.
Can Apheris be used without joining a federated network?
Yes. Teams can deploy and run models locally in their own environment for secure inference, benchmarking, and customization without participating in any federated data network. Networks are optional and extend — rather than replace — standalone functionality.
What types of drug discovery networks does Apheris support?
Apheris currently powers three federated networks: the AISB Network for protein-complex prediction, the ADMET Network for pharmacokinetics and drug metabolism in DMTA cycles, and the Antibody Developability Network for antibody R&D.
