Machine Learning
Fast training of support vector machines using sequential minimal optimization
Advances in kernel methods
The WEKA data mining software: an update
ACM SIGKDD Explorations Newsletter
Domain-domain interactions in obligate and non-obligate protein-protein interactions
BIBMW '11 Proceedings of the 2011 IEEE International Conference on Bioinformatics and Biomedicine Workshops
A model to predict and analyze protein-protein interaction types using electrostatic energies
BIBM '12 Proceedings of the 2012 IEEE International Conference on Bioinformatics and Biomedicine (BIBM)
| Hi-index | 0.00 |
The importance of protein-protein interactions (PPIs) is becoming increasingly appreciated, as these interactions lie at the core of virtually every biological process. Small molecule modulators that target PPIs are under exploration as new therapies. One of the greatest obstacles faced in crystallographically determining the 3D structures of proteins is coaxing the proteins to form ''artificial'' PPIs that lead to uniform crystals suitable for X-ray diffraction. This work compares interactions formed naturally, i.e., ''biological'', with those artificially formed under crystallization conditions or ''non-biological''. In particular, a detailed analysis of water molecules at the interfaces of high-resolution (@?2.30A) X-ray crystal structures of protein-protein complexes, where 140 are biological protein-protein complex structures and 112 include non-biological protein-protein interfaces, was carried out using modeling tools based on the HINT forcefield. Surprisingly few and relatively subtle differences were observed between the two types of interfaces: (i) non-biological interfaces are more polar than biological interfaces, yet there is better organized hydrogen bonding at the latter; (ii) biological associations rely more on water-mediated interactions with backbone atoms compared to non-biological associations; (iii) aromatic/planar residues play a larger role in biological associations with respect to water, and (iv) Lys has a particularly large role at non-biological interfaces. A support vector machines (SVMs) classifier using descriptors from this study was devised that was able to correctly classify 84% of the two interface types.