SHORT Case StudY

Reducing drug development costs through AI protein crystal detection.

Appsilon introduced Crystal Clear Vision (CCV), an innovative machine learning model that leverages transfer learning and Low-Rank Adaptation (LoRA) for fine-tuning, significantly outperforming MARCO.
BioPharma
Global top 1% Applied Science Institution, improving drug discovery processes.
Technologies used:
PyTorch
neptune.ai

What business problem
did we face?

The process of drug discovery is notoriously costly and time-consuming, with protein crystal identification being a critical step that traditionally requires substantial computational resources and expertise. The existing state-of-the-art model, MARCO, while effective, misses almost 10% of crystals and demands extensive computational effort. This inefficiency leads to missed opportunities in identifying potential drug candidates, directly impacting the speed and cost of bringing new medicines to market.

The solution we
proposed

Appsilon introduced Crystal Clear Vision (CCV), an innovative machine learning model that leverages transfer learning and Low-Rank Adaptation (LoRA) for fine-tuning, significantly outperforming MARCO.

By focusing on enhancing crystal detection accuracy and applying state-of-the-art techniques, CCV dramatically lowers the missed crystal rate to 3.6% (reduction in lost crystals of 61%). Thanks to this boost in efficiency, larger scale surveys are enabled, leading to a reduced time to discovering new drug candidates.

The impact of our solution
and its ROI

Crystal detection error rate reduced by 61% (from 9.3% originally to 3.6%).
Testimonial

“Every time you miss a protein crystal, because they are so rare, you risk missing on an important biomedical discovery.”

Patrick Charbonneau

Duke University, Lead Researcher in MARCO initiative

Explore Possibilities

Share Your Data Goals with Us

From advanced analytics to platform development and pharma consulting, we craft solutions tailored to your needs.

Talk to our Experts