Laboratory of Computational Systems Biology(RIKEN) Recruiting students for the academic year 2019

Associate Professor Kam Zhang(Kam Zhang)
E-mail: kamzhang{at}
Lab HP


【Key Words】Protein design; Protein structure prediction; X-ray ab initio phasing; Cryo-EM structure refinement; Computational drug design

 The complex biological functions of proteins are determined by their equally intricate three-dimensional structures. The correctly folded native structure is critical for the proper function of a protein in a cell. Small deviations from its native structure can often lead to malfunction of the protein and cause diseases. We are interested in understanding the protein functions through computational studies of their structures. Our research interests are on the following areas:

1. Protein folding and design

 Understanding the principles of protein folding especially the energetics will enable us to predict protein structures from their sequences. We have developed an efficient conformational sampling method for fragment-assembly based de novo protein structure prediction called EdaFold that uses an Estimation of Distribution Algorithm. This method has achieved top performance in the template-free modeling category of CASP10. We have developed one of the fastest exact cluster method called Durandal that can be used to identify a good protein model among many decoys.

 Protein design allows us to explore large regions of the protein universe not yet observed in nature. Recently, we have developed a very efficient method and have used it to design the first perfectly symmetric β propeller proteins that self-assemble according to simple arithmetic rules. We are interested in applying the protein design principles to create proteins with novel architectures, new biological functions or effective therapeutics.

2. AB initio phasing with de novo models

 We are developing new computational methods to solve the X-ray crystallographic phase problem for protein structure determination. Our efforts are focused on improving de novo models predicted computationally so that they can be used as templates for structure determination by molecular replacement. We have developed an error-estimation guided model rebuilding method that can efficiently improve de novo models with increased success rate for molecular replacement. Recently,we have developed a fragmentation and assembly method that can use low accuracy de novo models for ab initio phasing.

3. Virtual screening and drug design

 Drug discovery is a long and costly endeavor that involves many stages of multidisciplinary collaborations. Our effort focuses on using computational tools to identify initial hit compounds for a given protein target (lead discovery) and optimize them into potent lead compounds (lead optimization). We also develop novel methods for the identification of small molecule inhibitors. We collaborate with biologists to validate our identified hit compounds by experimental assays. We also collaborate with structural biologists to understand the binding mode of hit compounds for lead optimization.


  1. Jiang, X., et al., J. Chem. Inf. Model., 56, 2413-2420 (2016)
  2. Kumar, A., et al., J. Comput-Aided Mol. Des., 30, 457-469 (2016).
  3. Kumar, A., et al., J. Chem. Inf. Model., 56, 965-973 (2016).
  4. Voet, A., et al., Angew. Chem. Int. Ed., 54, 9857-9860 (2015).
  5. Shrestha, R., et al., Acta Cryst., D71, 304-312 (2015)
  6. Voet, A., et al., PNAS, 111, 15102-15107 (2014).
  7. Kumar, A., et al., J. Chem. Inf. Model., 54, 2784-2793 (2014).
  8. Berenger, F., et al., J. Cheminformatics, 6: 23 (2014)
  9. Shrestha, R. et al., Proteins, 82, 2240-2252 (2014).
  10. Berenger, F., et al., Bioinformatics, 27, 939-945 (2011).


The University of Tokyo
Graduate School of Frontier Sciences, The University of Tokyo

Page Top