Columbia University

Technology Ventures

Algorithm for detecting potential structural interactions among proteins

Technology #cu12266

Questions about this technology? Ask a Technology Manager

Download Printable PDF

Image Gallery
Barry Honig
Managed By
Ron Katz

This technology is an algorithm that leverages genomic and structural information to detect potential interactions between proteins.

Unmet Need: Identifying potential protein structural interactions from sequence information

The identification of pairs of interacting proteins is essential for understanding cell regulatory mechanisms. However, most known protein-protein interactions have been deduced from exhaustive and time-consuming experimental studies such as the yeast two-hybrid assay, affinity purification, and manual curation. Additionally, while computational approaches have been developed to infer protein-protein interactions, they have failed to yield reliable genome-wide information.

The Technology: Algorithm identifies potential protein-protein interactions without time-consuming experimental studies

This technology is an algorithm that combines known information on the structure and sequence of proteins to predict new protein-protein interactions. Nicknamed PrePPI (predicting protein-protein interactions), this technology compares protein sequences to create structure-based sequence alignments that can be used to create interaction models between each pair of proteins in a dataset. These predicted protein pairs are then scored by the algorithm to yield the probability that the predicted interaction pair is true. By predicting the most likely protein-protein interactions, this technology bypasses the need for exhaustive experimentation, allowing researchers to selectively validate predicted pairs.

A benchmark study revealed that PrePPI provides superior information to that obtained from structural and non-structural evidence alone.


  • Predicting protein-protein interactions
  • Tool for improved drug development
  • Research tool for determining protein interactomes


  • Improved protein-protein interaction predictions
  • Predicts interactions without time-consuming experimental studies
  • Outperforms related methods with lower false positive rate and higher true positive rate

Lead Inventor:

Barry Honig, Ph.D.

Related Publications:

Tech Ventures Reference:

  • IR CU12266

  • Licensing Contact: Ron Katz