< Magical Power of Quantum Mechanics

1. What software to use for QM calculations?

Gaussian, Gamess, and ORCA are popular ones for organic chemists. WuXi AppTec provides us with Spartan, which uses Q-Chem as the calculation engine.


Figure 1. Highly cited QM calculation engines in the scientific literature[1]

2. Which DFT and basis set do you use as standard method?

Currently we use Density Functional Theory wB97X-D and basis set 6-31G* as our standard method. It provides a balance of speed and accuracy our organic chemists need.


Figure 2. Too many DFT choices[2]

3. Any book will you recommend to learn practical QM?

We searched for practical QM textbooks which could be useful for organic chemists. There is none. As such, our colleagues felt the need to share what they learned in articles published in https://chemistry.wuxiapptec.com/qm and https://www.wuxibiology.com/magical-power-of-qm. They are collated into eBook you could download for free from the same links.


Figure 3. Our practical QM book for organic chemists

4. How do WuXi AppTec chemists learn QM?

We started from retrospective analyses of reactions we run every day, especially the challenging ones we couldn’t predict properly intuitively. We optimized our methods, developed SOP, and turned QM into a prospective tool. We discussed and taught one another what we learned. It is a continuous team effort.


 Figure 4. Learning Pyramid

5. What is the difference between QM and ML for reactivity predictions?

QM focuses on physical reality, could pin point uniqueness of each substrate, each reagent. ML focuses on data, statistical average. Currently QM offers better guidance when it is not intuitive easy to make correct prospective analyses and is more useful in retrospective analyses for divergence in reactivity. As QM parameters are incorporated into Machine Learning, we anticipate chemistry ML tools to continue to improve.


Figure 5. Relevant data driven decision. Differences in nature of data[3]

6. What is the lead time to become proficient in QM?

Literature calculations that are useful for practical analyses are very limited. It is a continuous efforts in WuXi AppTec to develop these methods. We delegate our QM analyses to chemists with appropriate levels of proficiency and encourage everyone to learn. As we invest our efforts to analyze reactions of interest, we become more proficient. QM calculation methods and computational power are evolving very fast, new reactions continue to be discovered, etc. Over time, we learn “What to calculate, How to calculate, How to interpret the results, and How to improve the methods.” This is lifelong learning.


Figure 6. Proficiency is relative. Quantum Mechanics give us Insight and Numbers

7. QM analysis is relatively slow, how your teams do them fast enough for prospective analysis for synthetic planning?

We established SOP for different analyses, fit for purpose ones. From orbital to electrostatic potential map analyses which take a few minutes, to reaction energy profiles, transition state calculations which could take a few hours. Accessible QM Databases with properties of a large number of organic molecules pre-calculated are tremendously useful.


Figure 7. Fit for purpose calculations are fast enough for prospective analyses

8. What computer system do you use for QM calculations? Are they affordable to synthetic laboratories?

QM calculations are CPU intensive, good to run them on desktops with ≥8 cores processors, which have become more affordable. The balance between computational vs time+labor+ reagent+disposal costs has changed significantly. QM calculation is becoming a valuable Resource Optimization/Green Chemistry tool.


Figure 8. Multiple-core processors speed up routine QM calculations for synthetic planning

9. If we have specific chemistry we need to analyze, who shall we contact?

For specific computational chemistry needs, please email IDSU_CS@wuxiapptec.com

This article is written and edited by Yongsheng Chen and John S. Wai.


[1] L. Talirz, L.M. Ghiringhelli, B. Smit, Living Journal of Computational Molecular Science2021, 3(1), 1483.

[2] a) R. Peverati, Int. J. Quantum Chem. 2021;121:e26379  b) L. Goerigk, N. Mehta, Aust. J. Chem. 2019, 72, 563.

[3] a) F. Strieth-Kalthoff, F. Sandfort, M.H.S. Segler, F. Glorius, Chem. Soc. Rev.2020, 49, 6154. b) Y.F. Guan, C.W. Coley, H.Y. Wu, A. Ranasunghe, E. Heid, T.J. Struble, L. Pattanaik, W.H. Green, K.F. Jensen, Chem. Sci.2021, 12, 2198.

[4] https://www.davidparkins.com

[5] a) Spartan Spectra and Physical Properties Database: https://www.wavefun.com  b) QM7-X: J. Hoja, L.M. Sandonas, B.G. Ernst, A. Vazquez-Mayagoitia, R.A. DiStasio Jr, A. Tkatchenko, Sci Data, 2021, 8, 43.