Optimization‐based framework for computer‐aided molecular design

Abstract

A new framework to automate, augment, and accelerate steps in computer-aided molecular design is presented. The problem is tackled in three stages: (1) composition design, (2) structure determination, and (3) extended design. Composition identification and structure determination are decoupled to achieve computational efficiency. Using approximate group-contribution methods in the first stage, molecular compositions that fit design targets are identified. In the second stage, isomer structures of solution compositions are determined systematically, and structure-based property corrections are used to refine the solution pool. In the final stage, the design is extended beyond the scope of group-contribution methods by using problem-specific property models. At each design stage, novel optimization models and graph theoretic algorithms generate a large and diverse pool of candidates using an assortment of property models. The wide applicability and computational efficiency of the proposed methodology are illustrated through three case studies.

Generic mathematical programming formulation and solution for computer-aided molecular design

“…Detailed information for the case studies can be found in the supplementary material. Design of other molecules (Samudra and Sahinidis, 2013;Harper et al, 2001;Karunanithi et al, 2005) such as refrigerant design and various types of solvents can be obtained from the authors.…”

“…This relaxation is justified by the fact that the average errors in first-order property estimation of the GC + model rarely exceed 10% (Samudra & Sahinidis, 2013). But it is not always easy for the users to find the appropriate relaxations.…”

“…That means the optimal point obtained from the decomposed method might be a local optima. Samudra and Sahinidis (2013) used property relaxation method to avoid this situation. That is, instead of property interval […”

“…Harper et al (1999) proposed a framework for CAMD method, where the predesign phase defines the basic needs, the design phase determines the feasible candidates (generates molecules and tests for desired properties) and the post-design phase performs higher level analysis of the molecular structure and the final selection of the product. Samudra and Sahinidis (2013) proposed a new optimization model using relaxed property targets and refined property targets with structural corrections. It is usually difficult to model and solve the MILP/MINLP problem with structure information considered due to the increased size of the mathematical problem and number of alternatives.…”

“…Harper et al (1999) used a generate and test approach to decompose the CAMD problem; selection of building blocks (functional groups), combination of groups into chemically feasible molecules, estimation of the specified set of properties for the generated molecules, selection as candidate compounds, and finally, determination of those that match the specified set of properties. Samudra and Sahinidis (2013) decomposed the problem into three design steps: composition design, structure design and extended design. In composition design, a large number of compositions (molecules composed of groups) matching relaxed design criteria based on first-order property estimates are determined.…”