Artificial Intelligence in Molecular Optimization: Current Paradigms and Future Frontiers

Molecular optimization plays a pivotal role in many domains since it holds promise for improving the properties of lead molecules. The advent of artificial intelligence (AI)-driven molecular optimization has revolutionized lead optimization workflows, which have significantly accelerated the development of drug candidates. However, AI models are also confronted with new challenges in practical molecular optimization, such as high-dimensional chemical space and data sparsity issues. This paper initially highlights the inherent benefits of molecular optimization in terms of optimizing the properties and maintaining the structural similarity of lead molecules, thereby highlighting its critical role in drug discovery. The next section systematically categorizes and analyzes existing AI-aided molecular optimization methods, comprising iterative search in discrete chemical space, end-to-end generation in continuous latent space, and iterative search in continuous latent space methods. Finally, we discuss the key challenges in AI-aided molecular optimization methods, including molecular representations, dataset selection, the properties to be optimized, and optimization algorithms, while proposing potential solutions and future research directions. In summary, this review provides a comprehensive analysis of existing representative AI-aided molecular optimization methods, thereby offering guidance for future research directions.

In silico modeling of targeted protein degradation

Targeted protein degradation (TPD) techniques, particularly proteolysis-targeting chimeras (PROTAC) and molecular glue degraders (MGD), have offered novel strategies in drug discovery. With rapid advancement of computer-aided drug design (CADD) and artificial intelligence-driven drug discovery (AIDD) in the biomedical field, a major focus has become how to effectively integrate these technologies into the TPD drug discovery pipeline to accelerate development, shorten timelines, and reduce costs. Currently, the main research directions for applying CADD and AIDD in TPD include: 1) ternary complex modeling; 2) linker generation; 3) strategies to predict degrader targets, activities and ADME/T properties; 4) In silico degrader design and discovery. Models developed in these areas play a crucial role in target identification, drug design, and optimization at various stages of the discovery process. However, the limited size and quality of datasets related to TPD present challenges, leaving room for further improvement in these models. TPD involves the complex ubiquitin-proteasome system, with numerous factors influencing outcomes. Most current models adopt a static perspective to interpret and predict relevant tasks. In the future, it may be necessary to shift toward dynamic approaches that better capture the intricate relationships among these components. Furthermore, incorporating new and diverse chemical spaces will enhance the precision design and application of TPD agents.

Transition metal phosphide/ molybdenum disulfide heterostructures towards advanced electrochemical energy storage: recent progress and challenges

Transition metal phosphide @ molybdenum disulfide (TMP@MoS2) heterostructures, consisting of TMP as the core main catalytic body and MoS2 as the outer shell, can solve the three major problems in the field of renewable energy storage and catalysis, such as lack of resources, cost factors, and low cycling stability. The heterostructures synergistically combine the excellent conductivity and electrochemical performance of transition metal phosphides with the structural robustness and catalytic activity of molybdenum disulfide, which holds great promise for clean energy. This review addresses the advantages of TMP@MoS2 materials and their synthesis methods-e.g., hydrothermal routes and chemical vapor deposition regarding scalability and cost. Their electrochemical energy storage and catalytic functions e.g., hydrogen and oxygen evolution reactions (HER and OER) are also extensively explored. Their potential within battery and supercapacitor technologies is also assessed against leading performance metrics. Challenges toward industry-scale scalability, longevity, and environmental sustainability are also addressed, as are optimization and large-scale deployment strategies.

Druggability Studies of Benzene Sulfonamide Substituted Diarylamide (E3) as a Novel Diuretic

Background/Objectives: Urea transporters (UTs) play an important role in the urine-concentrating mechanism and have been regarded as a novel drug target for developing salt-sparing diuretics. Our previous studies found that diarylamides 1H and 25a are specific UT inhibitors and have oral diuretic activity. However, these compounds necessitate further optimization and comprehensive druggability studies. Methods: The optimal compound was identified through structural optimization. Experiments were conducted to investigate its UT inhibitory activity and evaluate its diuretic effect. Furthermore, disease models were utilized to assess the compound's efficacy in treating hyponatremia. Pharmacokinetic studies were performed to examine its metabolic stability, and toxicity tests were conducted to evaluate its safety. Results: Based on the chemical structure of compound 25a, we synthesized a novel diarylamide compound, E3, by introducing a benzenesulfonamide group into its side chain. E3 exhibited dose-dependent inhibition of UT at the nanomolar level and demonstrated oral diuretic activity without causing electrolyte excretion disorders in both mice and rats. Experiments on UT-B-/- and UT-A1-/- mice indicated that E3 enhances the diuretic effect primarily by inhibiting UT-A1 more effectively than UT-B. Furthermore, E3 displayed good metabolic stability and favorable pharmacokinetic characteristics. E3 significantly ameliorated hyponatremia through diuresis in a rat model. Importantly, E3 did not induce acute oral toxicity, subacute oral toxicity, genotoxicity, or cardiotoxicity. Conclusions: Our study confirms that E3 exerts a diuretic effect by specifically inhibiting UTs and has good druggability, which offers potential for E3 to be developed into a new diuretic for the treatment of hyponatremia.

Discovery, design, and engineering of enzymes based on molecular retrobiosynthesis

Biosynthesis-a process utilizing biological systems to synthesize chemical compounds-has emerged as a revolutionary solution to 21st-century challenges due to its environmental sustainability, scalability, and high stereoselectivity and regioselectivity. Recent advancements in artificial intelligence (AI) are accelerating biosynthesis by enabling intelligent design, construction, and optimization of enzymatic reactions and biological systems. We first introduce the molecular retrosynthesis route planning in biochemical pathway design, including single-step retrosynthesis algorithms and AI-based chemical retrosynthesis route design tools. We highlight the advantages and challenges of large language models in addressing the sparsity of chemical data. Furthermore, we review enzyme discovery methods based on sequence and structure alignment techniques. Breakthroughs in AI-based structural prediction methods are expected to significantly improve the accuracy of enzyme discovery. We also summarize methods for de novo enzyme generation for nonnatural or orphan reactions, focusing on AI-based enzyme functional annotation and enzyme discovery techniques based on reaction or small molecule similarity. Turning to enzyme engineering, we discuss strategies to improve enzyme thermostability, solubility, and activity, as well as the applications of AI in these fields. The shift from traditional experiment-driven models to data-driven and computationally driven intelligent models is already underway. Finally, we present potential challenges and provide a perspective on future research directions. We envision expanded applications of biocatalysis in drug development, green chemistry, and complex molecule synthesis.

Database of Nonaqueous Proton-Conducting Materials

This work presents the assembly of 48 papers, representing 74 different compounds and blends, into a machine-readable database of nonaqueous proton-conducting materials. SMILES was used to encode the chemical structures of the molecules, and we tabulated the reported proton conductivity, proton diffusion coefficient, and material composition for a total of 3152 data points. The data spans a broad range of temperatures ranging from -70 to 260 °C. To explore this landscape of nonaqueous proton conductors, DFT was used to calculate the proton affinity of 18 unique proton carriers. The results were then compared to the activation energy derived from fitting experimental data to the Arrhenius equation. It was found that while the widely recognized positive correlation between the activation energy and proton affinity may hold among closely related molecules, this correlation does not necessarily apply across a broader range of molecules. This work serves as an example of the potential analyses that can be conducted using literature data combined with emerging research tools in computation and data science to address specific materials design problems.

Chemically Informed Deep Learning for Interpretable Radical Reaction Prediction

Organic radical reactions are crucial in many areas of chemistry, including synthetic, biological, and atmospheric chemistry. We develop a predictive framework based on the interaction of molecular orbitals that operates on mechanistic-level radical reactions. Given our chemistry-aware model, all predictions are provided with different levels of interpretability. Our models are trained and evaluated using the RMechDB database of radical reaction steps. Our model predicts the correct orbital interaction and products for 96% of the test reactions in RMechDB. By chaining these predictions, we perform a pathway search capable of identifying all intermediates and byproducts of a radical reaction. We test the pathway search on two classes of problems in atmospheric and polymerization chemistry. RMechRP is publicly available online at https://deeprxn.ics.uci.edu/rmechrp/.

A review of large language models and autonomous agents in chemistry

Large language models (LLMs) have emerged as powerful tools in chemistry, significantly impacting molecule design, property prediction, and synthesis optimization. This review highlights LLM capabilities in these domains and their potential to accelerate scientific discovery through automation. We also review LLM-based autonomous agents: LLMs with a broader set of tools to interact with their surrounding environment. These agents perform diverse tasks such as paper scraping, interfacing with automated laboratories, and synthesis planning. As agents are an emerging topic, we extend the scope of our review of agents beyond chemistry and discuss across any scientific domains. This review covers the recent history, current capabilities, and design of LLMs and autonomous agents, addressing specific challenges, opportunities, and future directions in chemistry. Key challenges include data quality and integration, model interpretability, and the need for standard benchmarks, while future directions point towards more sophisticated multi-modal agents and enhanced collaboration between agents and experimental methods. Due to the quick pace of this field, a repository has been built to keep track of the latest studies: https://github.com/ur-whitelab/LLMs-in-science.

Predicting Rate Constants of Hydrogen Abstraction Reactions between OH/HO2 and Alkanes by Machine Learning Models

The hydrogen abstraction reactions by small radicals from fuel molecules play an important role in the oxidation of fuels. However, experimental measurements and/or theoretical calculations of their rate constants under combustion conditions are very challenging due to their high reactivity. Machine learning offers a promising approach to predicting thermal rate constants. In this work, three machine learning methods, XGB, FNN, and XGB-FNN hybrid algorithms, were employed to train and predict the rate constants of the hydrogen abstraction reactions between alkanes and OH/HO2. Six descriptors were selected according to the Pearson correlation coefficients, the importance of descriptors, and the clustering heat map. It was proven that the XGB-FNN model is the most robust. The constructed XGB-FNN model achieved an average deviation of 89.13% for the alkanes + OH reactions and 190.93% for the alkanes + HO2 reactions on their respective prediction sets. The model was also used to predict the rate constants of the reactions involving larger alkanes, demonstrating its extrapolation capability. Furthermore, the model has the ability to distinguish the reactivity of the reactions with the hydrogen atom abstracted at different sites of alkane.

Generation of Rational Drug-like Molecular Structures Through a Multiple-Objective Reinforcement Learning Framework

As an appealing approach for discovering novel leads, the key advantage of de novo drug design lies in its ability to explore a much broader dimension of chemical space, without being confined to the knowledge of existing compounds. So far, many generative models have been described in the literature, which have completely redefined the concept of de novo drug design. However, many of them lack practical value for real-world drug discovery. In this work, we have developed a graph-based generative model within a reinforcement learning framework, namely, METEOR (Molecular Exploration Through multiplE-Objective Reinforcement). The backend agent of METEOR is based on the well-established GCPN model. To ensure the overall quality of the generated molecular graphs, we implemented a set of rules to identify and exclude undesired substructures. Importantly, METEOR is designed to conduct multi-objective optimization, i.e., simultaneously optimizing binding affinity, drug-likeness, and synthetic accessibility of the generated molecules under the guidance of a special reward function. We demonstrate in a specific test case that without prior knowledge of true binders to the chosen target protein, METEOR generated molecules with superior properties compared to those in the ZINC 250k data set. In conclusion, we have demonstrated the potential of METEOR as a practical tool for generating rational drug-like molecules in the early phase of drug discovery.

Hit me with your best shot: Integrated hit discovery for the next generation of drug targets

Identification of high-quality hit chemical matter is of vital importance to the success of drug discovery campaigns. However, this goal is becoming ever harder to achieve as the targets entering the portfolios of pharmaceutical and biotechnology companies are increasingly trending towards novel and traditionally challenging to drug. This demand has fuelled the development and adoption of numerous new screening approaches, whereby the contemporary hit identification toolbox comprises a growing number of orthogonal and complementary technologies including high-throughput screening, fragment-based ligand design, affinity screening (affinity-selection mass spectrometry, differential scanning fluorimetry, DNA-encoded library screening), as well as increasingly sophisticated computational predictive approaches. Herein we describe how an integrated strategy for hit discovery, whereby multiple hit identification techniques are tactically applied, selected in the context of target suitability and resource priority, represents an optimal and often essential approach to maximise the likelihood of identifying quality starting points from which to develop the next generation of medicines.

Understanding predictions of drug profiles using explainable machine learning models

PURPOSE

The analysis of absorption, distribution, metabolism, and excretion (ADME) molecular properties is of relevance to drug design, as they directly influence the drug's effectiveness at its target location. This study concerns their prediction, using explainable Machine Learning (ML) models. The aim of the study is to find which molecular features are relevant to the prediction of the different ADME properties and measure their impact on the predictive model.

METHODS

The relative relevance of individual features for ADME activity is gauged by estimating feature importance in ML models' predictions. Feature importance is calculated using feature permutation and the individual impact of features is measured by SHAP additive explanations.

RESULTS

The study reveals the relevance of specific molecular descriptors for each ADME property and quantifies their impact on the ADME property prediction.

CONCLUSION

The reported research illustrates how explainable ML models can provide detailed insights about the individual contributions of molecular features to the final prediction of an ADME property, as an effort to support experts in the process of drug candidate selection through a better understanding of the impact of molecular features.

An extensive review on lung cancer therapeutics using machine learning techniques: state-of-the-art and perspectives

There are over 100 types of human cancer, accounting for millions of deaths every year. Lung cancer alone claims over 1.8 million lives per year and is expected to surpass 3.2 million by 2050, which underscores the urgent need for rapid drug development and repurposing initiatives. The application of AI emerges as a pivotal solution to developing anti-cancer therapeutics. This state-of-the-art review aims to explore the various applications of AI in lung cancer therapeutics. Predictive models can analyse large datasets, including clinical data, genetic information, and treatment outcomes, for novel drug design and to generate personalised treatment recommendations, potentially optimising therapeutic strategies, enhancing treatment efficacy, and minimising adverse effects. A thorough literature review study was conducted based on articles indexed in PubMed and Scopus. We compiled the use of various machine learning approaches, including CNN, RNN, GAN, VAEs, and other AI techniques, enhancing efficiency with accuracy exceeding 95%, which is validated through a computer-aided drug design process. AI can revolutionise lung cancer therapeutics, streamlining processes and saving biological scientists' time and effort-however, further research is needed to overcome challenges and fully unlock AI's potential in Lung Cancer Therapeutics.

Exploring the Intersection of Geophysics and Diagnostic Imaging in the Health Sciences

To develop diagnostic imaging approaches, this paper emphasizes the transformational potential of merging geophysics with health sciences. Diagnostic imaging technology improvements have transformed the health sciences by enabling earlier and more precise disease identification, individualized therapy, and improved patient care. This review article examines the connection between geophysics and diagnostic imaging in the field of health sciences. Geophysics, which is typically used to explore Earth's subsurface, has provided new uses of its methodology in the medical field, providing innovative solutions to pressing medical problems. The article examines the different geophysical techniques like electrical imaging, seismic imaging, and geophysics and their corresponding imaging techniques used in health sciences like tomography, magnetic resonance imaging, ultrasound imaging, etc. The examination includes the description, similarities, differences, and challenges associated with these techniques and how modified geophysical techniques can be used in imaging methods in health sciences. Examining the progression of each method from geophysics to medical imaging and its contributions to illness diagnosis, treatment planning, and monitoring are highlighted. Also, the utilization of geophysical data analysis techniques like signal processing and inversion techniques in image processing in health sciences has been briefly explained, along with different mathematical and computational tools in geophysics and how they can be implemented for image processing in health sciences. The key findings include the development of machine learning and artificial intelligence in geophysics-driven medical imaging, demonstrating the revolutionary effects of data-driven methods on precision, speed, and predictive modeling.

Advances in the Application of Artificial Intelligence-Based Spectral Data Interpretation: A Perspective

The interpretation of spectral data, including mass, nuclear magnetic resonance, infrared, and ultraviolet-visible spectra, is critical for obtaining molecular structural information. The development of advanced sensing technology has multiplied the amount of available spectral data. Chemical experts must use basic principles corresponding to the spectral information generated by molecular fragments and functional groups. This is a time-consuming process that requires a solid professional knowledge base. In recent years, the rapid development of computer science and its applications in cheminformatics and the emergence of computer-aided expert systems have greatly reduced the difficulty in analyzing large quantities of data. For expert systems, however, the problem-solving strategy must be known in advance or extracted by human experts and translated into algorithms. Gratifyingly, the development of artificial intelligence (AI) methods has shown great promise for solving such problems. Traditional algorithms, including the latest neural network algorithms, have shown great potential for both extracting useful information and processing massive quantities of data. This Perspective highlights recent innovations covering all of the emerging AI-based spectral interpretation techniques. In addition, the main limitations and current obstacles are presented, and the corresponding directions for further research are proposed. Moreover, this Perspective gives the authors' personal outlook on the development and future applications of spectral interpretation.

Machine Learning Methods for Small Data Challenges in Molecular Science

Small data are often used in scientific and engineering research due to the presence of various constraints, such as time, cost, ethics, privacy, security, and technical limitations in data acquisition. However, big data have been the focus for the past decade, small data and their challenges have received little attention, even though they are technically more severe in machine learning (ML) and deep learning (DL) studies. Overall, the small data challenge is often compounded by issues, such as data diversity, imputation, noise, imbalance, and high-dimensionality. Fortunately, the current big data era is characterized by technological breakthroughs in ML, DL, and artificial intelligence (AI), which enable data-driven scientific discovery, and many advanced ML and DL technologies developed for big data have inadvertently provided solutions for small data problems. As a result, significant progress has been made in ML and DL for small data challenges in the past decade. In this review, we summarize and analyze several emerging potential solutions to small data challenges in molecular science, including chemical and biological sciences. We review both basic machine learning algorithms, such as linear regression, logistic regression (LR), k-nearest neighbor (KNN), support vector machine (SVM), kernel learning (KL), random forest (RF), and gradient boosting trees (GBT), and more advanced techniques, including artificial neural network (ANN), convolutional neural network (CNN), U-Net, graph neural network (GNN), Generative Adversarial Network (GAN), long short-term memory (LSTM), autoencoder, transformer, transfer learning, active learning, graph-based semi-supervised learning, combining deep learning with traditional machine learning, and physical model-based data augmentation. We also briefly discuss the latest advances in these methods. Finally, we conclude the survey with a discussion of promising trends in small data challenges in molecular science.

Identifying potential drug-target interactions based on ensemble deep learning

Introduction

Drug-target interaction prediction is one important step in drug research and development. Experimental methods are time consuming and laborious.

Methods

In this study, we developed a novel DTI prediction method called EnGDD by combining initial feature acquisition, dimensional reduction, and DTI classification based on Gradient boosting neural network, Deep neural network, and Deep Forest.

Results

EnGDD was compared with seven stat-of-the-art DTI prediction methods (BLM-NII, NRLMF, WNNGIP, NEDTP, DTi2Vec, RoFDT, and MolTrans) on the nuclear receptor, GPCR, ion channel, and enzyme datasets under cross validations on drugs, targets, and drug-target pairs, respectively. EnGDD computed the best recall, accuracy, F1-score, AUC, and AUPR under the majority of conditions, demonstrating its powerful DTI identification performance. EnGDD predicted that D00182 and hsa2099, D07871 and hsa1813, DB00599 and hsa2562, D00002 and hsa10935 have a higher interaction probabilities among unknown drug-target pairs and may be potential DTIs on the four datasets, respectively. In particular, D00002 (Nadide) was identified to interact with hsa10935 (Mitochondrial peroxiredoxin3) whose up-regulation might be used to treat neurodegenerative diseases. Finally, EnGDD was used to find possible drug targets for Parkinson's disease and Alzheimer's disease after confirming its DTI identification performance. The results show that D01277, D04641, and D08969 may be applied to the treatment of Parkinson's disease through targeting hsa1813 (dopamine receptor D2) and D02173, D02558, and D03822 may be the clues of treatment for patients with Alzheimer's disease through targeting hsa5743 (prostaglandinendoperoxide synthase 2). The above prediction results need further biomedical validation.

Discussion

We anticipate that our proposed EnGDD model can help discover potential therapeutic clues for various diseases including neurodegenerative diseases.

CMGN: a conditional molecular generation net to design target-specific molecules with desired properties

The rational design of chemical entities with desired properties for a specific target is a long-standing challenge in drug design. Generative neural networks have emerged as a powerful approach to sample novel molecules with specific properties, termed as inverse drug design. However, generating molecules with biological activity against certain targets and predefined drug properties still remains challenging. Here, we propose a conditional molecular generation net (CMGN), the backbone of which is a bidirectional and autoregressive transformer. CMGN applies large-scale pretraining for molecular understanding and navigates the chemical space for specified targets by fine-tuning with corresponding datasets. Additionally, fragments and properties were trained to recover molecules to learn the structure-properties relationships. Our model crisscrosses the chemical space for specific targets and properties that control fragment-growth processes. Case studies demonstrated the advantages and utility of our model in fragment-to-lead processes and multi-objective lead optimization. The results presented in this paper illustrate that CMGN has the potential to accelerate the drug discovery process.

Designing hexaphyrins for high-potential NLO switches: the synergy of core-modifications and meso-substitutions

Due to the enormous size of the chemical compound space, usually only small regions are traversed with traditional direct molecular design approaches making the discovery for novel functionalized molecules for nonlinear optical applications challenging. By applying inverse molecular design algorithms, we aim to efficiently explore larger regions of the compound space in search of promising hexaphyrin-based molecular switches as measured by their first-hyperpolarizability (β_HRS) contrast. We focus on the 28R → 30R switch with a functionalization pattern allowing for centrosymmetric OFF states yielding zero β_HRS response. This switch is particularly challenging as full meso-substitution with a single type of functional group or core-modifications result in almost no contrast enhancement. We carried out four inverse design procedures during which two sets of core-modifications and three sets of meso-substitutions sites were systematically optimized. All 4 optimal switches are characterized by a mix of meso-substitutions and core-modifications, of which the best performing switch yields a 10-fold improvement over the parent macrocycle. Throughout the inverse design procedures, we collected and analyzed a database biased towards high NLO contrasts that contains 277 different patterns for hexaphyrin-based switches. We derived three design rules to obtain highly functional 28R → 30R NLO switches: (I) a combination of 2 strong EWG and 1 EDG group is the ideal recipe for increasing the NLO contrast, though their position also plays an important role. (II) The type of core-modification is less important when only the diagonal positions are core-modified. Switches with 4 core-modifications show a clear preference for oxygen. (III) Keeping centrosymmetry in the OFF state remains highly beneficial given the investigated functionalization pattern. Finally, we have demonstrated that combining meso-substitutions with core-modifications can synergistically improve the NLO contrast.

Conditional Molecular Generation Net Enables Automated Structure Elucidation Based on 13C NMR Spectra and Prior Knowledge

Structure elucidation of unknown compounds based on nuclear magnetic resonance (NMR) remains a challenging problem in both synthetic organic and natural product chemistry. Library matching has been an efficient method to assist structure elucidation. However, it is limited by the coverage of libraries. In addition, prior knowledge such as molecular fragments is neglected. To solve the problem, we propose a conditional molecular generation net (CMGNet) to allow input of multiple sources of information. CMGNet not only uses ¹³C NMR spectrum data as input but molecular formulas and fragments of molecules are also employed as input conditions. Our model applies large-scale pretraining for molecular understanding and fine-tuning on two NMR spectral data sets of different granularity levels to accommodate structure elucidation tasks. CMGNet generates structures based on ¹³C NMR data, molecular formula, and fragment information, with a recovery rate of 94.17% in the top 10 recommendations. In addition, the generative model performed well in the generation of various classes of compounds and in the structural revision task. CMGNet has a deep understanding of molecular connectivities from ¹³C NMR, molecular formula, and fragments, paving the way for a new paradigm of deep learning-assisted inverse problem-solving.

Computational Design and Manufacturing of Sustainable Materials through First-Principles and Materiomics

Engineered materials are ubiquitous throughout society and are critical to the development of modern technology, yet many current material systems are inexorably tied to widespread deterioration of ecological processes. Next-generation material systems can address goals of environmental sustainability by providing alternatives to fossil fuel-based materials and by reducing destructive extraction processes, energy costs, and accumulation of solid waste. However, development of sustainable materials faces several key challenges including investigation, processing, and architecting of new feedstocks that are often relatively mechanically weak, complex, and difficult to characterize or standardize. In this review paper, we outline a framework for examining sustainability in material systems and discuss how recent developments in modeling, machine learning, and other computational tools can aid the discovery of novel sustainable materials. We consider these through the lens of materiomics, an approach that considers material systems holistically by incorporating perspectives of all relevant scales, beginning with first-principles approaches and extending through the macroscale to consider sustainable material design from the bottom-up. We follow with an examination of how computational methods are currently applied to select examples of sustainable material development, with particular emphasis on bioinspired and biobased materials, and conclude with perspectives on opportunities and open challenges.

Computer-aided multi-objective optimization in small molecule discovery

Molecular discovery is a multi-objective optimization problem that requires identifying a molecule or set of molecules that balance multiple, often competing, properties. Multi-objective molecular design is commonly addressed by combining properties of interest into a single objective function using scalarization, which imposes assumptions about relative importance and uncovers little about the trade-offs between objectives. In contrast to scalarization, Pareto optimization does not require knowledge of relative importance and reveals the trade-offs between objectives. However, it introduces additional considerations in algorithm design. In this review, we describe pool-based and de novo generative approaches to multi-objective molecular discovery with a focus on Pareto optimization algorithms. We show how pool-based molecular discovery is a relatively direct extension of multi-objective Bayesian optimization and how the plethora of different generative models extend from single-objective to multi-objective optimization in similar ways using non-dominated sorting in the reward function (reinforcement learning) or to select molecules for retraining (distribution learning) or propagation (genetic algorithms). Finally, we discuss some remaining challenges and opportunities in the field, emphasizing the opportunity to adopt Bayesian optimization techniques into multi-objective de novo design.

Deep Reinforcement Learning for Molecular Inverse Problem of Nuclear Magnetic Resonance Spectra to Molecular Structure

Spectroscopy is the study of how matter interacts with electromagnetic radiation. The spectra of any molecule are highly information-rich, yet the inverse relation of spectra to the corresponding molecular structure is still an unsolved problem. Nuclear magnetic resonance (NMR) spectroscopy is one such critical technique in the scientists' toolkit to characterize molecules. In this work, a novel machine learning framework is proposed that attempts to solve this inverse problem by navigating the chemical space to find the correct structure given an NMR spectra. The proposed framework uses a combination of online Monte Carlo tree search (MCTS) and a set of graph convolution networks to build a molecule iteratively. Our method can predict the structure of the molecule ∼80% of the time in its top 3 guesses for molecules with <10 heavy atoms. We believe that the proposed framework is a significant step in solving the inverse design problem of NMR spectra.