We are ACE Team who solve chemistry problems using computers!

​

We are trying to solve 3 main problems by computation using deep learning and physical chemistry.

We are developing deep learning algorithms for smart drug discovery. 

​

​

Developers

          Seongok Ryu, Jaechang Lim, Sang-Yeon Hwang, Jeong-eun Park

​

Hit discovery

• Structure-based deep learning models for predicting drug-target interaction (J. Chem. Inf. Model. 2019)

• Activity-based deep learning models

• Phenomic-based deep learning models

Hit-to-Lead

• ADME/T prediction

• Enhancing the potency of drug candidates using molecular generative models (J. Cheminfo. 2018, J. Chem. Inf. Model. 2020, Chem. Sci. 2020)

• Off-target selectivity prediction

​

Reliable use of deep neural networks

• A Bayesian deep learning for uncertainty quantification of prediction results (Chem. Sci. 2019)

​

Architecture of deep neural network for predicting molecular properties

• Augmented graph convolutional networks (arXiv:1805.10988 (JCIM, submitted))

We are developing automated tools for reaction prediction and retrosynthesis based on quantum calculation tools and deep learning

​

Main works

​

We are developing a platform for designing materials that are important in commercials using deep learning and quantum calculation

Nanoparticles

• Fuel cell with Prof. Sang Woo Han's group (KAIST): ACS Appl Mater Interfaces 4, 6228 (2012)

• Catalytic reaction on the surface with Prof. Sang Woo Han's group (KAIST): Chem Comm 50, 9454 (2014), Chem Comm 8, 2450 (2016)

• Simulating quantum plasmon with Prof. Seol Ryu, Sci. Rep. 7, 15775 (2017)

​

Mechanism study of organic reactions

• CO2 conversion study

• Ligand design with Prof. Hyunwoo Kim's group (KAIST): OL 16, 5490 (2014)

• Cross-coupling reaction with Hyunwoo Kim's group (KAIST): OL 18, 616 (2016)

​

Polymers and Biomaterials

• Designing transparent polymers for electronic applications with Prof. Sang Yul Kim's group (KAIST): Sci Adv, 4, eaau1956 (2018)

• Non-covalent interactions between biomolecules with Prof. Haeshin Lee's group (KAIST): Adv Funct Mater 22, 4711 (2012)

• Non-covalent interactions between biomolecules with Prof. Kwang S Kim's group (UNIST): JCTC 9, 2090 (2013)

​

Two-dimensional carbon allotropes

• Design of graphdiyne structure using the concept of aromaticity: Carbon 98, 404 (2016)

• Ion transfer mechanism of Lithium cation in the single-layer graphdiyne: ACS Appl Mater Interfaces, 11, 2677 (2019)

​

Molecular electronics and spintronics

• DNA sequencing device: Nat Nanotech  6, 162 (2011)

• Supermagnetoresistance: Nat Nanotech  4, 408 (2008)

• Porphyrin-based spin filtering device: JACS 133, 9364 (2011)

• Molecular magnet: Angew Chem Int Ed  52, 3389 (2012)

We are developing the quantum chemistry software, namely ACE-Molecule. It uses a numerical grid and Lagrange-Sinc functions as a basis set and supports the following calculations.

Developers

Sunghwan Choi (KISTI), Kwangwoo Hong, Jaewook Kim, Seongok Ryu, Sang-Yeon Hwang, Sungwoo Kang, Jaechang Lim, Hyeonsu Kim, Jeheon Woo, Yongjun Kim and Woo Youn Kim

​​

Basics of KS-DFT

•  Single-point KS calculations for both closed and open shell systems: JCP 142, 094116 (2015)

•  Density/potential mixing: Broyden, Pulay, DIIS

•  Acceleration of initial guess and SCF convergence: IJQC 116, 1397 (2016)

• Geometry optimization and supersampling method: JCP 114, 094141 (2016)

• Implementation of PAW and its accuracy:  IJQC 116, 644 (2016)

​​

Functionals

• All LDA/GGA functionals in libxc

• KLI approximated functional for the exact exchange potential: BKCS 36, 998 (2015)

• Hybrid functionals with local exchange potential (PBE0, B3LYP): PCCP 19, 10177 (2017)

• Long-range corrected hybrid functionals with local exchange potential (LC-wPBE): CPC 230, 21 (2018)

​​

Excited state calculations

• TDDFT with LDA, GGA, and hybrid: PCCP 19, 10177 (2017), CPC 230, 21 (2018)

• CIS with KS orbitals: PCCP 17, 31434 (2015)

• CISD with KS orbitals: JCP 145, 224309 (2016)

• Simulation of plasmonic excitations using DDA and ADA: Sci. Rep. 7, 15775 (2017)

• Simulation of optical excitation of the hydrogen passivated Si QD using the long-range corrected hybrid functional: CPC 230, 21 (2018)

Parallelization

• MPI, OpenMP, GPU, MIC, and heterogeneous: JCC 37, 2193 (2016)

• Diagonalization with Lanzos method: submitted to IJQC 118, e225622 (2018)

​

Official Homepage

For more information about ACE-Molecule, visit the official website: https://gitlab.com/aceteam.kaist/ACE-Molecule

We are developing a software for automated exploration of reaction paths. Mathematical methods such as linear algebra and graph theory are combined with chemical rules for high speed computations yet with reliable accuracy. Combinatorial searching for all chemically possible intermediates from given reactants and products is quickly performed, resulting in namely the reaction network. Then, graph-theoretic analysis determines most favorable reaction paths. We are planning to extend this software for the prediction of reaction products together with all kinetic information.

​

​

​

​

​

​

​

​

​

​

​

​

​

​

Developers

Yeonjoon Kim, Sunghwan Choi, Jin Woo Kim, Zeehyo Kim, and Woo Youn Kim

​

Intermediate sampling

• Basin-Hopping Monte Carlo simulation: JCTC 10, 2419 (2014)

• Searching for isomers: Carbon 98, 404 (2016)

• Enumerating molecular graphs: Chem. Sci. (2017)

Structure conversion method

• Automated structure conversion from adjacency matrix to 3D geometry: BKCS 36, 1769 (2015)

​

Reaction network analysis

• Application of Dijkstra and Yen's shortest path algorithm : JCTC 10, 2419 (2014)

• Time propagation through direct solving rate equations: under development: JPCA 2019.