Energy functions in de novo protein design: current challenges and future prospects.

By Zhixiu Li, Yuedong Yang, Jian Zhan, Liang Dai, Yaoqi Zhou. 2013

78 citations recorded by the source corpus.

Subfields: Materials Chemistry, Molecular Biology

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Materials Chemistry

1. Yaoqi Zhou
2. Yuedong Yang
3. David Baker
4. Zhixiu Li
5. Kuldip K. Paliwal
6. Jian Zhan
7. Jack Hanson
8. Brian Kuhlman

In Molecular Biology

1. Yaoqi Zhou
2. Yuedong Yang
3. Kuldip K. Paliwal
4. Jack Hanson
5. David Baker
6. Zhixiu Li
7. Jian Zhan
8. Tanja Kortemme

Citations in the corpus

• ProDCoNN: Protein design using a convolutional neural network (2019, background): “17,24 The average sequence recovery rates achieved by the current top-performing protein-design programs are around 35% on a small number of proteins.”
• Computational Protein Design with Deep Learning Neural Networks (2018, background): “25,26 In many cases, the sequences from computational design are subject to further filtering by considering various factors such as shape complementarity 9 and in silico folding free energy landscape, 27,28 where human experience and familiarity with the designed protein play an important role,…”
• SPIN2: Predicting sequence profiles from protein structures using deep neural networks (2018, background): “11 However, existing energy functions for protein design are not yet sufficiently accurate, 11,12 although recent progress has been made.”
• Protein Engineering by Combined Computational and In Vitro Evolution Approaches. (2016, background): “In CPD calculations, the protein backbone is either fixed or allowed small movements [71] and side chain–side chain and side chain–backbone interactions are evaluated using an atomicbased energy function representing physical interactions, such as van der Waals energy, electrostatic interactions, hy”
• Direct prediction of profiles of sequences compatible with a protein structure by neural networks with fragment‐based local and energy‐based nonlocal profiles (2014, background): “Consistent with other methods 17,18 , sequence identities between predicted and actual sequences are lower for proteins with higher fraction of surface residues.”
• Protein design with a comprehensive statistical energy function and boosted by experimental selection for foldability (2014, background): “However, success rates of automated design with common targets has remained low 14,15 .”
• Rapid Sampling of Hydrogen Bond Networks for Computational Protein Design (2018, methodology): “The energy functions that are typically used for computational protein design 22,23 , including the Rosetta Energy Function 24 , are expressed as the sum of pairwise energies, which is important for computational efficiency and algorithmic compatibility 25 .”