Apheris is transforming the AI data challenge in the life sciences by employing federated computing, a method that shows considerable potential for improving both privacy and efficiency in machine learning.
In the life sciences field, AI is heavily dependent on data; however, privacy and regulatory issues significantly limit the use of health data. This challenge has been a major obstacle, with approximately 97% of all healthcare data remaining entirely unused because it is isolated within organisations and cannot be shared.
Highlights
How Apheris Tackles This Challenge
Apheris has introduced a solution that utilises federated computing, which allows data to be utilised securely for AI model training without the need to transfer the data physically. Here’s an overview of the process:
- Local Computation: Computations are conducted locally, ensuring that sensitive data stays within the organisation.
- Secure Data Collaboration: Only essential outputs are gathered centrally, safeguarding the sensitive information of data owners.
- Cross-Institutional Collaboration: This strategy promotes cooperation among various organisations, including pharmaceutical firms and hospitals, without compromising data privacy.
Recent Developments and Funding
Apheris has recently secured a Series A funding round of $8.25 million, spearheaded by DeepTech investors OTB Ventures and eCAPITAL. This investment will enable Apheris to broaden its influence and establish itself as a premier provider of federated computing solutions within the life sciences sector. The funding will be utilised to develop the most extensive and secure life science data network by linking distributed health and life sciences data for analytics and AI.
Industry Impact and Partnerships
Apheris has already attracted significant interest from prominent clients, such as Roche and various hospitals. The company is also a member of the AI Structural Biology (AISB) Consortium, which comprises leading pharmaceutical companies like AbbVie, Boehringer Ingelheim, Johnson & Johnson, and Sanofi. This consortium utilises Apheris’ secure federated learning framework to tailor AI drug discovery models using proprietary pharmaceutical data while maintaining data privacy.
Overcoming Technical Challenges
While federated computing presents numerous advantages, it also introduces specific challenges:
- Data and Model Security: There is a potential risk of sensitive information being exposed through exchanged parameters during the learning process. Apheris counters this with stringent cybersecurity measures.
- Data Heterogeneity: Different clients might have distinct data distributions and computational resources, which can cause issues like global model drift. Apheris is developing strategies to manage these variances.
- Communication Overhead: As the number of nodes involved increases, communication overhead can become substantial. Apheris is optimising both local training iteration times and global aggregation algorithms to mitigate this overhead.
Apheris is indeed reimagining the AI data bottleneck in life sciences through federated computing, enabling the full potential of AI and machine learning in the sector while preserving data privacy and security.