Geometric Projections of Conformational Space

This work concerns the development of novel projections of a protein's conformational space. Our original approach in FeLTr employs geometric coordinates based on the Ultrafast Shape recognition coordinates. A 3-dimensional grid is employed in FeLTr to project the explored conformational space.

Our work leverages this low-dimensional projection to efficiently cluster sampled conformations. Additionally, clustering is employed to collapse geometrically-similar conformations and so maintain only a small representative sample of the explored conformational space in memory. Maintaining this reduced ensemble significantly lowers the memory footprint of FeLTr. The figure to the right illustrates an example of the reduction in memory that we are able to achieve. Results show that the reduced ensemble size allows the algorithm to explore larger conformational spaces and more effectively sample conformations near a protein's native structure.

In subsequent work we investigate different projection coordinates based on contact matrices. Essentially, intra-molecular distances between atoms in a conformation are used to build a profile of the conformation's internal geometry. An eigendecomposition of the resulting contact matrix elucidates the essential connectivity of the conformation in the first principal component. Local Hashing is then employed to project FeLTr-computed conformations onto the space of essential connectivity vectors and so obtain a lower-dimensional embedding of the explored conformational space.

In recent work we extend the usage of projections to guide the search towards a given structural state. The application for this line of work is computation of motions connecting two given functional states of a protein. A progress coordinate is defined between the states, and this coordinate is used to group conformations. This information is employed by the search to steer the exploration towards the goal state.

This work appears in: 1) Kevin Molloy and Amarda Shehu "Elucidating the Ensemble of Functionally-relevant Transitions in Protein Systems with a Robotics-inspired Method." BMC Structural Biology J, 2013; 2) Kevin Molloy and Amarda Shehu "A Robotics-inspired Method to Sample Conformational Paths Connecting Known Functionally-relevant Structures in Protein Systems." IEEE BIBMW - Comput Struct Biol Workshop (CSBW), pg. 56-63, Philadelphia, PA, 2012." 3) Brian Olson, Seyed-Farid Hendi, and Amarda Shehu "Investigating the Role of Projections in Guiding Protein Conformational Search" J Bioinf and Comput Biol; 4) Brian Olson, Seyed-Farid Hendi, and Amarda Shehu "Protein Conformational Search with Geometric Projections" Computational Structural Biology Workshop (CSBW) at IEEE International Conference on Bioinformatics and Biomedicine (IEEE BIBM), Atlanta, GA, 2011; 5) Brian Olson, Kevin Molloy, and Amarda Shehu "In Search of the Protein Native State with a Probabilistic Sampling Approach" J Bioinf and Comp Biol 2011, 9(3):383-398; and 6) Brian Olson, Kevin Molloy, and Amarda Shehu "Enhancing Sampling of the Conformational Space Near the Protein Native State" Intl. Conference on Bio-inspired Models of Network, Information, and Computing Systems (BIONETICS),Boston, MA, 2010.

This is Kevin's presentation at CSBW 2012 on this line of work for computing protein motions.

On this Project:

  • Kevin Molloy<

    Brian Olson (Ph.D. alumni)

    Seyed Hendi (M.S. alumni)

    Amarda Shehu

This material is based upon work supported by the National Science Foundation under Grant No. 1016995 and IIS CAREER Award No. 1144106. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation.