Knowledge Engineering in Chemistry: From Expert Systems to Agents of Creation.

Passing knowledge from human to human is a natural process that has continued since the beginning of humankind. Over the past few decades, we have witnessed that knowledge is no longer passed only between humans but also from humans to machines. The latter form of knowledge transfer represents a cornerstone in artificial intelligence (AI) and lays the foundation for knowledge engineering (KE). In order to pass knowledge to machines, humans need to structure, formalize, and make knowledge machine-readable. Subsequently, humans also need to develop software that emulates their decision-making process. In order to engineer chemical knowledge, chemists are often required to challenge their understanding of chemistry and thinking processes, which may help improve the structure of chemical knowledge.Knowledge engineering in chemistry dates from the development of expert systems that emulated the thinking process of analytical and organic chemists. Since then, many different expert systems employing rather limited knowledge bases have been developed, solving problems in retrosynthesis, analytical chemistry, chemical risk assessment, etc. However, toward the end of the 20th century, the AI winters slowed down the development of expert systems for chemistry. At the same time, the increasing complexity of chemical research, alongside the limitations of the available computing tools, made it difficult for many chemistry expert systems to keep pace.In the past two decades, the semantic web, the popularization of object-oriented programming, and the increase in computational power have revitalized knowledge engineering. Knowledge formalization through ontologies has become commonplace, triggering the subsequent development of knowledge graphs and cognitive software agents. These tools enable the possibility of interoperability, enabling the representation of more complex systems, inference capabilities, and the synthesis of new knowledge.This Account introduces the history, the core principles of KE, and its applications within the broad realm of chemical research and engineering. In this regard, we first discuss how chemical knowledge is formalized and how a chemist's cognition can be emulated with the help of reasoning algorithms. Following this, we discuss various applications of knowledge graph and agent technology used to solve problems in chemistry related to molecular engineering, chemical mechanisms, multiscale modeling, automation of calculations and experiments, and chemist-machine interactions. These developments are discussed in the context of a universal and dynamic knowledge ecosystem, referred to as The World Avatar (TWA).

Automated Calibration of a Poly(oxymethylene) Dimethyl Ether Oxidation Mechanism Using the Knowledge Graph Technology.

In this paper, we develop a knowledge graph-based framework for the automated calibration of combustion reaction mechanisms and demonstrate its effectiveness on a case study of poly(oxymethylene)dimethyl ether (PODEn, where n = 3) oxidation. We develop an ontological representation for combustion experiments, OntoChemExp, that allows for the semantic enrichment of experiments within the J-Park simulator (JPS, theworldavatar.com), an existing cross-domain knowledge graph. OntoChemExp is fully capable of supporting experimental results in the Process Informatics Model (PrIMe) database. Following this, a set of software agents are developed to perform experimental result retrieval, sensitivity analysis, and calibration tasks. The sensitivity analysis agent is used for both generic sensitivity analyses and reaction selection for subsequent calibration. The calibration process is performed as a sampling task, followed by an optimization task. The agents are designed for use with generic models but are demonstrated with ignition delay time and laminar flame speed simulations. We find that calibration times are reduced, while accuracy is increased compared to manual calibration, achieving a 79% decrease in the objective function value, as defined in this study. Further, we demonstrate how this workflow is implemented as an extension of the JPS.

Cost-effectiveness of niraparib, rucaparib, and olaparib for treatment of platinum-resistant, recurrent ovarian carcinoma.

BACKGROUND: Olaparib was approved on December 19, 2014 by the US FDA as 4th-line therapy (and beyond) for patients with germline BRCA1/2 mutations; rucaparib was approved on December 19, 2016 as 3rd-line therapy (and beyond) for germline or somatic BRCA1/2-mutated recurrent disease. On October 23, 2019, niraparib was approved for treatment of women with damaging mutations in BRCA1/2 or other homologous recombination repair genes who had been treated with three or more prior regimens. We compared the cost-effectiveness of PARPi(s) with intravenous regimens for platinum-resistant disease. METHODS: Median progression-free survival (PFS) and toxicity data from regulatory trials were incorporated in a model which transitioned patients through response, hematologic complications, non-hematologic complications, progression, and death. Using TreeAge Pro 2017, each PARPi(s) was compared separately to non­platinum-based and bevacizumab-containing regimens. Costs of IV drugs, managing toxicities, infusions, and supportive care were estimated using 2017 Medicare data. Incremental cost-effectiveness ratios (ICERs) were calculated and PFS was reported in quality adjusted life months for platinum-resistant populations. RESULTS: Non­platinum-based intravenous chemotherapy was most cost effective ($6,412/PFS-month) compared with bevacizumab-containing regimens ($12,187/PFS-month), niraparib ($18,970/PFS-month), olaparib ($16,327/PFS-month), and rucaparib ($16,637/PFS-month). ICERs for PARPi(s) were 3-3.5× times greater than intravenous non­platinum-based regimens. CONCLUSION: High costs of orally administered PARPi(s) were not mitigated or balanced by costs of infusion and managing toxicities of intravenous regimens typically associated with lower response and shorter median PFS. Balancing modest clinical benefit with costs of novel therapies remains problematic and could widen disparities among those with limited access to care.

A Markov model to evaluate cost-effectiveness of antiangiogenesis therapy using bevacizumab in advanced cervical cancer.

OBJECTIVE: To evaluate the cost-effectiveness of bevacizumab in recurrent/persistent and metastatic cervical cancer using recently reported updated survival and toxicology data. METHODS: A Markov decision tree based on the Gynecologic Oncology Group 240 randomized trial was created. The 2013 MediCare Services Drug Payment Table and Physician Fee Schedule provided costs. In the 5-year model subjects transitioned through the following states: response, progression, minor complications, severe complications, and death. Patients experiencing a health utility per month according to treatment effectiveness were calculated. Because cervical cancer survival is measured in months rather than years, results were reported in both quality adjusted cervical cancer life months and years (QALmonth, QALY), adjusted from a baseline of having advanced cervical cancer during a month. RESULTS: The estimated total cost of therapy with bevacizumab is approximately 13.2 times that for chemotherapy alone, adding $73,791 per 3.5months (0.29year) of life gained, resulting in an incremental cost-effectiveness ratio (ICER) of $21.083 per month of added life. The ICER increased to $5775 per month of added life and $24,597/QALmonth ($295,164/QALY) due to the smaller difference in QALmonths. With 75% bevacizumab cost reduction, the ICER is $6737/QALmonth ($80,844/QALY), which translates to $23,580 for the 3.5month (0.29year) gain in OS. CONCLUSIONS: Increased costs are primarily related to the cost of drug and not the management of bevacizumab-induced complications. Cost reductions in bevacizumab result in dramatic declines in the ICER, suggesting that cost reconciliation in advanced cervical cancer may be possible through the availability of biosimilars, and/or less expensive, equally efficacious anti-angiogenesis agents.

A dynamic knowledge graph approach to distributed self-driving laboratories.

The ability to integrate resources and share knowledge across organisations empowers scientists to expedite the scientific discovery process. This is especially crucial in addressing emerging global challenges that require global solutions. In this work, we develop an architecture for distributed self-driving laboratories within The World Avatar project, which seeks to create an all-encompassing digital twin based on a dynamic knowledge graph. We employ ontologies to capture data and material flows in design-make-test-analyse cycles, utilising autonomous agents as executable knowledge components to carry out the experimentation workflow. Data provenance is recorded to ensure its findability, accessibility, interoperability, and reusability. We demonstrate the practical application of our framework by linking two robots in Cambridge and Singapore for a collaborative closed-loop optimisation for a pharmaceutically-relevant aldol condensation reaction in real-time. The knowledge graph autonomously evolves toward the scientist's research goals, with the two robots effectively generating a Pareto front for cost-yield optimisation in three days.

The digital lab manager: Automating research support.

Laboratory management automation is essential for achieving interoperability in the domain of experimental research and accelerating scientific discovery. The integration of resources and the sharing of knowledge across organisations enable scientific discoveries to be accelerated by increasing the productivity of laboratories, optimising funding efficiency, and addressing emerging global challenges. This paper presents a novel framework for digitalising and automating the administration of research laboratories through The World Avatar, an all-encompassing dynamic knowledge graph. This Digital Laboratory Framework serves as a flexible tool, enabling users to efficiently leverage data from diverse systems and formats without being confined to a specific software or protocol. Establishing dedicated ontologies and agents and combining them with technologies such as QR codes, RFID tags, and mobile apps, enabled us to develop modular applications that tackle some key challenges related to lab management. Here, we showcase an automated tracking and intervention system for explosive chemicals as well as an easy-to-use mobile application for asset management and information retrieval. Implementing these, we have achieved semantic linking of BIM and BMS data with laboratory inventory and chemical knowledge. Our approach can capture the crucial data points and reduce inventory processing time. All data provenance is recorded following the FAIR principles, ensuring its accessibility and interoperability.

From Platform to Knowledge Graph: Evolution of Laboratory Automation.

High-fidelity computer-aided experimentation is becoming more accessible with the development of computing power and artificial intelligence tools. The advancement of experimental hardware also empowers researchers to reach a level of accuracy that was not possible in the past. Marching toward the next generation of self-driving laboratories, the orchestration of both resources lies at the focal point of autonomous discovery in chemical science. To achieve such a goal, algorithmically accessible data representations and standardized communication protocols are indispensable. In this perspective, we recategorize the recently introduced approach based on Materials Acceleration Platforms into five functional components and discuss recent case studies that focus on the data representation and exchange scheme between different components. Emerging technologies for interoperable data representation and multi-agent systems are also discussed with their recent applications in chemical automation. We hypothesize that knowledge graph technology, orchestrating semantic web technologies and multi-agent systems, will be the driving force to bring data to knowledge, evolving our way of automating the laboratory.

Predicting Power Conversion Efficiency of Organic Photovoltaics: Models and Data Analysis.

In this paper, the ability of three selected machine learning neural and baseline models in predicting the power conversion efficiency (PCE) of organic photovoltaics (OPVs) using molecular structure information as an input is assessed. The bidirectional long short-term memory (gFSI/BiLSTM), attentive fingerprints (attentive FP), and simple graph neural networks (simple GNN) as well as baseline support vector regression (SVR), random forests (RF), and high-dimensional model representation (HDMR) methods are trained to both the large and computational Harvard clean energy project database (CEPDB) and the much smaller experimental Harvard organic photovoltaic 15 dataset (HOPV15). It was found that the neural-based models generally performed better on the computational dataset with the attentive FP model reaching a state-of-the-art performance with the test set mean squared error of 0.071. The experimental dataset proved much harder to fit, with all of the models exhibiting a rather poor performance. Contrary to the computational dataset, the baseline models were found to perform better than the neural models. To improve the ability of machine learning models to predict PCEs for OPVs, either better computational results that correlate well with experiments or more experimental data at well-controlled conditions are likely required.