Persistent Homology for the Quantitative Analysis of the Structure and Stability of Carboranes

Chen Dong,Zhang Ming-Zheng,Chen Hai-Biao,Xie Zuo-Wei,Wei Guo-Wei,Pan Feng
DOI: https://doi.org/10.14102/j.cnki.0254-5861.2011-2889
IF: 0.847
2020-01-01
Chinese journal of structural chemistry
Abstract:Persistent homology is a powerful and novel tool for quantifying the inherent topological features of structure. In this work, we used the persistent homology for the first time to study the closo-carboranes C2Bn-2Hn (n = 5 similar to 20) and their parent structures closo-boranes dianions BnHn2- (n = 5 similar to 20), where multiple elements are present. All these structures are first investigated with the standard Vitoris-Rips complex. We interpret all barcodes representation and associate them with structural details. By means of average bar length, a linear regression model was established to construct the relationship between persistent homology features and molecular stability, which was expressed by the relative energies. For closo-boranes dianions, we only use B atom set since B and H atoms are in pairs. The average lengths of beta(0), beta(1) and beta(2) bars are used as the features for linear regression, and excellent correlation coefficient (0.977) between the values predicted by persistent homology and those by quantum calculations was achieved. For closo-carboranes, C-B atom set (ignore the differences in the atoms), B atom set and C atom set were considered to get the persistent homology features (since there were only two C atoms in C2Bn-2Hn, only beta(0) bars were considered), and seven average bar lengths were calculated, respectively. Pearson coefficient of 0.937 was obtained. We found that the stability of carboranes showed a high linear correlation with the characteristics generated from topological bars in H-0, H-1 and H-2. The results show that the topological information generated by persistent homology can be extended and applied to multi-element systems.
What problem does this paper attempt to address?