About
Protein crystallization has long been a time and labor-intensive bottleneck in protein structure determination. Over the last few decades, the adoption of laboratory automation, including liquid handlers and imagers, has enabled scientists to focus more on experimental design, greatly increasing their throughput and resulting in significantly more data. This happened at a cost of more time investment to evaluate images of the crystallization experiments. Formulatrix developed AI-based methods to fully automate drop scoring and crystal identification to alleviate this burden.
Initially, integrating a scientist-led and Google-developed auto-scoring model, MARCO, into Rock Maker was reasonably practical. Still, it struggled to identify crystals mixed with other classes in a drop. To improve this, Formulatrix developed Sherlock, an advanced AI auto-scoring model trained on larger, more diverse datasets. It is equipped with new features and the ability to identify crystals present with some other classes in a drop. These features make Sherlock the model of choice for Rock Maker users.