Scholarlink

<p>Atomic connectivity group contribution (ACGC) is a method developed by atomic adjacent group (AAG), shape factors and atomic connectivity factors (ACFs) for property prediction. As a crucial parameter for dealing with the challenge of accurately predicting properties of isomers, ACFs are defined for describing the global position of each group in a molecule. In this work, ACFs plus (ACFs+) is proposed to describe the local position of a group by considering the contribution of the core atom in AAG and nearby atoms. As such, ACGC plus (ACGC+) models are developed with ACFs+ to predict key phase transition properties of organic compounds (i.e., Δ<sub>fus</sub><i>S</i>, Δ<sub>vap</sub><i>H</i>°, Δ<sub>sub</sub><i>H</i>°, Δ<sub>fus</sub><i>H, T</i><sub>b</sub>, <i>T</i><sub>m</sub>, <i>T</i><sub>c</sub>, <i>P</i><sub>c</sub>, and <i>V</i><sub>c</sub>). Both predictability and robustness are rigorously validated using external validation and cross-validation. The <i>R</i><sub>test</sub><sup>2</sup> values for phase transition entropy and enthalpy range from 0.906 to 0.992, the <i>R</i><sub>test</sub><sup>2</sup> values for critical properties are greater than 0.989, and the <i>R</i><sub>test</sub><sup>2</sup> values for <i>T</i><sub>b</sub> and <i>T</i><sub>m</sub> are 0.979 and 0.845, respectively, which indicate high predictability of ACGC + models. The <i>R</i><sup>2</sup> values for all properties are close to the <i>R</i><sub>train</sub><sup>2</sup> values, which further validates the stability of the ACGC+ model. Furthermore, the mean absolute errors of the ACGC+ models decreased by 1.44%–7.91% compared to the ACGC models. These results demonstrate that the ACGC+ method provides high accuracy in predicting the properties of phase transitions.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

ACGC plus: a general group contribution framework for diverse properties prediction

  • 期刊论文

  • 发表日期：

    2026年5月25日

查看全文
  • Xian Liang,
  • Yuai Zheng,
  • Yuan Sun,
  • Songhan Zhang,
  • Jingxuan Xue,
  • Li-Hong Lin,
  • Qiaoyan Shang,
  • Yin-Ning Zhou,
  • Fangyou Yan

摘要

Atomic connectivity group contribution (ACGC) is a method developed by atomic adjacent group (AAG), shape factors and atomic connectivity factors (ACFs) for property prediction. As a crucial parameter for dealing with the challenge of accurately predicting properties of isomers, ACFs are defined for describing the global position of each group in a molecule. In this work, ACFs plus (ACFs+) is proposed to describe the local position of a group by considering the contribution of the core atom in AAG and nearby atoms. As such, ACGC plus (ACGC+) models are developed with ACFs+ to predict key phase transition properties of organic compounds (i.e., ΔfusS, ΔvapH°, ΔsubH°, ΔfusH, Tb, Tm, Tc, Pc, and Vc). Both predictability and robustness are rigorously validated using external validation and cross-validation. The Rtest2 values for phase transition entropy and enthalpy range from 0.906 to 0.992, the Rtest2 values for critical properties are greater than 0.989, and the Rtest2 values for Tb and Tm are 0.979 and 0.845, respectively, which indicate high predictability of ACGC + models. The R2 values for all properties are close to the Rtrain2 values, which further validates the stability of the ACGC+ model. Furthermore, the mean absolute errors of the ACGC+ models decreased by 1.44%–7.91% compared to the ACGC models. These results demonstrate that the ACGC+ method provides high accuracy in predicting the properties of phase transitions.