Dr. Yaron Orenstein's Lab


The Orenstein Lab at Ben-Gurion University of the Negev develops algorithms to infer predictive models of molecular interactions based on high-throughput biological data. The lab was established in October 2017 and is led by Dr. Yaron Orenstein. Dr. Orenstein completed his BSc, MSc, and PhD at Tel-Aviv University in Computer Science and Electrical Engineering. He was trained as a post-doc at the Computer Science and Artificial Intelligence Laboratory at MIT, and spent a semester as a research fellow at the Simons Institute for Theory of Computing at UC Berkeley. Since his PhD, his research area is Bioinformatics.

Partner's Genome Editing vision‚Äč

The consortium for us is an exciting opportunity to work on a very high-impact project and tackle new computational challenges that arise from the unique data of the CRISPR/CAS9 system. We are always thrilled to analyze huge molecular datasets, i.e. millions of DNA or RNA sequences, each with a measured molecular phenotype, and infer a model to predict the phenotype (in the consortium case, whether CAS9 cuts at or edits a specific position). Moreover, as the field of deep learning is evolving fast and providing better tools by the day, we hope to apply the latest developments in the field, such as reinforcement learning for bioinformatics sequence design problems, to get the best results in the experiments that will be performed as part of the consortium.

Partner's activity within the consortium

First, we apply deep neural networks on the unique high-throughput data that will be collected by the participating labs in the consortium. These neural networks enable off-target and on-target predictions for a specific guide RNA, similarly to modeling specificity of specific RNA- and DNA-binding proteins. Moreover, we visualize the key features of the networks on the sequence level to better understand the mechanism of each guide RNA. Furthermore, we develop an algorithm to predict an optimal guide RNA, i.e. high sensitivity and specificity, for a specific target set. Second, we design the sequence libraries for these high-throughput experiments to maximize the relevant information that can be measured in one experiment. The computational challenge is to generate the smallest library since experimental resources are limited. On top of that, since these libraries are generated by template oligos, which may contain degenerate nucleotides, we can even generate compact template libraries under different biological or experimental constraints, to cover all desired sequences to be measured against the CRISPR/CAS9 system. Purchasing several template oligos to generate the complete library may be much cheaper than purchasing an oligo for each sequence in the library.