An Application of Computer‐Aided Molecular Design (CAMD) Using the Signature Molecular Descriptor—Part 1. Identification of Surface Tension Reducing Agents and the Search for Shrinkage Reducing Admixtures

Abstract

The development of new admixtures for concrete is normally an experimental endeavor in that the molecular scaffolds of existing admixtures are modified and tested. This approach is time consuming, incremental and typically expensive. Alternatively, a computer‐aided molecular design (CAMD) approach is proposed that uses the Signature molecular descriptor. CAMD is the application of computer‐implemented algorithms that are utilized to design molecules with optimally predicted properties such that they can be tested and evaluated for efficacy. The property of interest here is the surface tension of compounds in aqueous solutions as this property is related to shrinkage in concrete. In particular, we have chosen two classes of compounds, amines and glycol ethers, as they present opportunities for use as shrinkage reducing admixtures (SRAs). By evaluating the initial surface tension reduction in these compounds in solution with water, a number of structure–property conjectures associated with the effect of these compounds were developed. From these conjectures, 14 compounds were identified and utilized as a training set for the CAMD of new compounds. After creating and refining a quantitative structure–property relationship (QSPR) model for surface tension reduction, a structure enumeration algorithm was employed to generate structures outside of the original training set that have optimally predicted properties. In work, the CAMD approach is introduced as well as the identification of new compounds with the greatest predicted impact on the surface tension reduction in water. Furthermore, the surface tension reduction for the newly identified compounds was experimentally evaluated.

Pharmaceutical Machine Learning: Virtual High-Throughput Screens Identifying Promising and Economical Small Molecule Inhibitors of Complement Factor C1s

“…It should be noted that the pipeline correlates structural feature patterns in compounds with experimental data and applies those correlations, in the form of models, to find new potential ligands. It is not equipped to identify why specific ligands are biologically active while others are not, though speculations can be drawn by correlating atomic Signature to model coefficients as was done in prior work [ 79 ].…”

Cements in the 21st century: Challenges, perspectives, and opportunities

“…In an effort to streamline and accelerate the discovery of admixture molecules, Kayello et al . and Shlonimskaya et al .…”

“…Some applications of ML to cementitious materials and related chemical admixtures have been made already, and they provide a glimpse of the latent power and flexibility of such methods for designing cementitious composites when sufficient training data are available. Phase segmentation of cement powder microstructures is now routinely accomplished by unsupervised ML, using clustering algorithms trained on subsets of 2D backscattered electron micrographs and X‐ray microanalysis data of polished cement powder cross sections .…”

Behavior of organic compounds in the pore solution of cement—experimental and molecular dynamics study

“…Thus, surface tension reduction of pore water is the controlling factor for shrinkage mitigation . However, most studies looked at surface tension reduction of SRas in pure water, whereas, few have considered pore solution …”

“…Kayello et al (2014) recently demonstrated a Computer‐Aided Molecular Design (CAMD) method using an inverse‐quantitative structure property relationship (I‐QSPR) approach to identify new SRas. They found that glycol ether acetate compounds have shrinkage reducing capacity and have less adverse effects on compressive strength . Glycol ether acetate compounds, thus, appear to be a new class of SRa that may biodegrade more readily than some current commercial SRA compounds and retain early age strength.…”