Improving chemical reaction yield prediction using pre-trained graph neural networks.

Graph neural networks (GNNs) have proven to be effective in the prediction of chemical reaction yields. However, their performance tends to deteriorate when they are trained using an insufficient training dataset in terms of quantity or diversity. A promising solution to alleviate this issue is to pre-train a GNN on a large-scale molecular database. In this study, we investigate the effectiveness of GNN pre-training in chemical reaction yield prediction. We present a novel GNN pre-training method for performance improvement.Given a molecular database consisting of a large number of molecules, we calculate molecular descriptors for each molecule and reduce the dimensionality of these descriptors by applying principal component analysis. We define a pre-text task by assigning a vector of principal component scores as the pseudo-label to each molecule in the database. A GNN is then pre-trained to perform the pre-text task of predicting the pseudo-label for the input molecule. For chemical reaction yield prediction, a prediction model is initialized using the pre-trained GNN and then fine-tuned with the training dataset containing chemical reactions and their yields. We demonstrate the effectiveness of the proposed method through experimental evaluation on benchmark datasets.

Data-driven autonomous operation of VOCs removal system.

Removal of volatile organic compounds (VOCs) from the air has been an important issue in many industrial fields. Traditionally, the operation of VOCs removal systems has relied on fixed operating conditions determined by domain experts based on their expertise and intuition. In practice, this manual operation cannot respond immediately to changes in the system environment. To facilitate the autonomous operation of the system, the operating conditions should be optimized properly in real time to adapt to the changes in the system environment. Recently, optimization frameworks have been widely applied to real-world industrial systems across various domains using different approaches. The primary motivation for this study is the effective implementation of an optimization framework targeting a VOCs removal system. In this paper, we present a data-driven autonomous operation method for optimizing the operating conditions of a VOCs removal system to enhance the overall performance. An optimization problem is formulated with the decision variables denoting the parameters associated with the operating condition, the environmental variables representing the measurements for the system environment, the constraints specifying the control ranges of the parameters, and the objective function representing the system performance as determined by the operating conditions and environment. Using the previous operation data from the system, a neural network is trained to model the system performance as a function of the decision and environmental variables to approximate the objective function. For the current state of the system environment, the optimal operating condition is derived by solving the optimization problem. A case study of a targeted VOCs removal system demonstrates that the proposed method effectively optimizes the operating conditions for improved system performance without intervention from domain experts.

Retention Time Prediction through Learning from a Small Training Data Set with a Pretrained Graph Neural Network.

Graph neural networks (GNNs) have shown remarkable performance in predicting the retention time (RT) for small molecules. However, the training data set for a particular target chromatographic system tends to exhibit scarcity, which poses a challenge because the experimental process for measuring RT is costly. To address this challenge, transfer learning has been used to leverage an abundant training data set from a related source task. In this study, we present an improved transfer learning method to better predict the RT of molecules for a target chromatographic system by learning from a small training data set with a pretrained GNN. We use a graph isomorphism network as the architecture of the GNN. The GNN is pretrained on the METLIN-SMRT data set and is then fine-tuned on the target training data set for a fixed number of training iterations using the limited-memory Broyden-Fletcher-Goldfarb-Shanno optimizer with a learning rate decay. We demonstrate that the proposed method achieves superior predictive performance on various chromatographic systems compared with that of the existing transfer learning methods, especially when only a small training data set is available for use. A potential avenue for future research is to leverage multiple small training data sets from different chromatographic systems to further enhance the generalization performance.

Generative Modeling to Predict Multiple Suitable Conditions for Chemical Reactions.

In synthesis planning, it is important to determine suitable reaction conditions such that a chemical reaction proceeds as intended. Recent research attempts based on machine learning have proven to be effective in recommending reaction elements for specific categories regarding critical chemical context and operating conditions. However, existing methods can only make a single prediction per reaction and do not directly provide a complete specification of the reaction elements as the prediction. Therefore, their achievable performance is limited. In this study, we propose a generative modeling approach to predict multiple different reaction conditions for a chemical reaction, each of which fully specifies critical reaction elements such that these elements can be directly used as a feasible reaction condition. We formulate the problem of predicting reaction conditions as sampling from a generative distribution. We model the distribution by introducing a variational autoencoder augmented with a graph neural network and learn it from a reaction dataset. For a query reaction, multiple predictions can be obtained by repeated sampling from the distribution. Through experimental investigation on the reaction datasets of four major types of cross-coupling reactions, we demonstrate that the proposed method significantly outperforms existing methods in retrieving ground-truth reaction conditions.

Scalable graph neural network for NMR chemical shift prediction.

Graph neural networks (GNNs) have been proven effective in the fast and accurate prediction of nuclear magnetic resonance (NMR) chemical shifts of a molecule. Existing methods, despite their effectiveness, suffer from high space complexity and are therefore limited to relatively small molecules. In this work, we propose a scalable GNN for NMR chemical shift prediction. To reduce the space complexity, we sparsify the graph representation of a molecule by regarding only heavy atoms as nodes and their chemical bonds as edges. To better learn from the sparsified graph representation, we improve the message passing and readout functions in the GNN. For the message passing function, we adapt the attention mechanism and residual connection to better capture local information around each node. For the readout function, we use both node-level and graph-level embeddings as the local and global information to better predict node-level chemical shifts. Through the experimental investigation using 13C and 1H NMR datasets, we demonstrate that the proposed method yields higher prediction accuracy and is more scalable to large molecules having many heavy atoms.

Predicting perioperative outcomes of robot-assisted radical cystectomy: Data from the Asian Robot-Assisted Radical Cystectomy Consortium.

OBJECTIVES: To report the perioperative outcomes of robot-assisted radical cystectomy and elucidate their risk factors. METHODS: A review of the Asian Robot-Assisted Radical Cystectomy Consortium database from 2007 to 2020 was performed. The perioperative outcomes studied included complication rates, time to solid food intake, estimated blood loss, length of hospital stay, and 30-day readmission rates. RESULTS: Of 568 patients, the overall complication rate was 49.2%, comprising major complications in 15.6%. Preoperative hydronephrosis was associated with an increased risk of major complications (odds ratio 3.27, 95% confidence interval 1.48-7.26, P = 0.004) while neoadjuvant chemotherapy was protective (odds ratio 0.46, 95% confidence interval 0.25-0.84, P = 0.012). The median time to solid food intake was 4 days (interquartile range 3-7) and smoking was a risk factor (odds ratio 4.28, 95% confidence interval 2.36-7.79, P < 0.001) for prolonged time to solid food intake. Median length of hospital stay was 13 days (interquartile range 9-19), and diabetes mellitus (odds ratio 1.66, 95% confidence interval 1.08-2.56, P = 0.021), neoadjuvant chemotherapy (odds ratio 2.21, 95% confidence interval 1.46-3.33, P < 0.001), and orthotopic bladder substitute creation (odds ratio 2.82, 95% confidence interval 1.90-4.18, P < 0.001) were independent risk factors for prolonged length of hospital stay. The 30-day readmission rate was 23.4% and higher in those with bilateral hydronephrosis (odds ratio 4.58, 95% confidence interval 1.97-10.65, P < 0.001) and orthotopic bladder substitute creation (odds ratio 1.87, 95% confidence interval 1.16-3.02, P = 0.010). CONCLUSIONS: There are preoperative conditions which are significant risk factors for adverse perioperative outcomes in robot-assisted radical cystectomy. Most are potentially modifiable and can direct strategies to reduce surgical morbidity related to this major oncological procedure.

Uncertainty-aware prediction of chemical reaction yields with graph neural networks.

In this paper, we present a data-driven method for the uncertainty-aware prediction of chemical reaction yields. The reactants and products in a chemical reaction are represented as a set of molecular graphs. The predictive distribution of the yield is modeled as a graph neural network that directly processes a set of graphs with permutation invariance. Uncertainty-aware learning and inference are applied to the model to make accurate predictions and to evaluate their uncertainty. We demonstrate the effectiveness of the proposed method on benchmark datasets with various settings. Compared to the existing methods, the proposed method improves the prediction and uncertainty quantification performance in most settings.

Molecular search by NMR spectrum based on evaluation of matching between spectrum and molecule.

Inferring molecular structures from experimentally measured nuclear magnetic resonance (NMR) spectra is an important task in many chemistry applications. Herein, we present a novel method implementing an automated molecular search by NMR spectrum. Given a query spectrum and a pool of candidate molecules, the matching score of each candidate molecule with respect to the query spectrum is evaluated by introducing a molecule-to-spectrum estimation procedure. The candidate molecule with the highest matching score is selected. This procedure does not require any prior knowledge of the corresponding molecular structure nor laborious manual efforts by chemists. We demonstrate the effectiveness of the proposed method on molecular search using 13C NMR spectra.

Evolutionary design of molecules based on deep learning and a genetic algorithm.

Evolutionary design has gained significant attention as a useful tool to accelerate the design process by automatically modifying molecular structures to obtain molecules with the target properties. However, its methodology presents a practical challenge-devising a way in which to rapidly evolve molecules while maintaining their chemical validity. In this study, we address this limitation by developing an evolutionary design method. The method employs deep learning models to extract the inherent knowledge from a database of materials and is used to effectively guide the evolutionary design. In the proposed method, the Morgan fingerprint vectors of seed molecules are evolved using the techniques of mutation and crossover within the genetic algorithm. Then, a recurrent neural network is used to reconstruct the final fingerprints into actual molecular structures while maintaining their chemical validity. The use of deep neural network models to predict the properties of these molecules enabled more versatile and efficient molecular evaluations to be conducted by using the proposed method repeatedly. Four design tasks were performed to modify the light-absorbing wavelengths of organic molecules from the PubChem library.

Perioperative Outcomes of Robot-Assisted Radical Cystectomy with Intracorporeal Versus Extracorporeal Urinary Diversion.

PURPOSE: This study was designed to investigate and compare the perioperative outcomes of intracorporeal urinary diversion (ICUD) versus extracorporeal urinary diversion (ECUD) following robotic-assisted radical cystectomy (RARC) in patients with localized bladder cancer from the Asian Robot-Assisted Radical Cystectomy (RARC) Consortium. METHODS: The Asian RARC registry was a multicenter registry involving nine centers in Asia. Consecutive patients who underwent RARC were included. Patient and disease characteristics, intraoperative details, and perioperative outcomes were reviewed and compared between the ICUD and ECUD groups. Postoperative complications were the primary outcomes, whereas secondary outcomes were the estimated blood loss and the duration of hospitalization. Multivariate regression analyses were performed to adjust potential confounders. RESULTS: From 2007 to 2020, 556 patients underwent RARC; 55.2% and 44.8% had ICUD and ECUD, respectively. ICUD group had less estimated blood loss (423.1 ± 361.1 vs. 541.3 ± 474.3 mL, p = 0.002) and a shorter hospital stay (15.7 ± 12.3 vs 17.8 ± 11.6 days, p = 0.042) than the ECUD group. Overall complication rates were similar between the two groups. Upon multivariate analysis, ICUD was associated with less estimated blood loss (Regression coefficient: - 143.06, 95% confidence interval [CI]: - 229.60 to - 56.52, p = 0.001) and a shorter hospital stay (Regression coefficient: - 2.37, 95% CI: - 4.69 to - 0.05, p = 0.046). In addition, ICUD was not associated with any increased risks of minor, major, and overall complications. CONCLUSIONS: RARC with ICUD was safe and technically feasible with similar postoperative complication rates as ECUD, with additional benefits of reduced blood loss and a shorter hospitalization.

Predictive Modeling of NMR Chemical Shifts without Using Atomic-Level Annotations.

Recently, machine learning has been successfully applied to the prediction of nuclear magnetic resonance (NMR) chemical shifts. To build a prediction model, the existing methods require a training data set that comprises molecules whose NMR-active atoms are annotated with their chemical shifts. However, the laborious task of atomic-level annotation must be manually conducted by chemists. Thus, it becomes difficult to perform large-scale annotation. To address this issue, we propose a weakly supervised learning method to enable the predictive modeling of NMR chemical shifts without requiring explicit atomic-level annotations in the training data set. For the training data set, the proposed method only requires the annotation of chemical shifts at the molecular level. As a prediction model, we build a message passing neural network (MPNN) that predicts the chemical shifts of individual NMR-active atoms in a molecule. Using a loss function that is invariant to the permutation of atoms in a molecule, the model is trained in a weakly supervised manner to minimize the molecular-level difference between a set of predicted chemical shifts and the corresponding set of actual chemical shifts across the training data set. Accordingly, during the training, the chemical shifts predicted by the model are approximately aligned with the actual chemical shifts in a data-driven fashion. The proposed method performs comparably to the existing fully supervised methods in terms of predicting the chemical shifts of 1H and 13C NMR spectra for small molecules.

Neural Message Passing for NMR Chemical Shift Prediction.

Fast and accurate prediction of NMR spectra enables automatic structure validation and elucidation of molecules on a large scale. In this Article, we propose an improved method of learning from an NMR database to predict the chemical shifts of NMR-active atoms of a new molecule. For this purpose, we use a message passing neural network that operates on the graph representation of a molecule. The compactness and informativeness of the graph representation are enhanced by treating hydrogen atoms implicitly and incorporating various node and edge features. Experimental investigation demonstrates that the proposed method achieves higher prediction performance for the chemical shifts in the 1H NMR and 13C NMR spectra of small molecules. We apply this method to determine the correct molecular structure for a new NMR spectrum by searching from a set of candidate molecules.

Compressed graph representation for scalable molecular graph generation.

Recently, deep learning has been successfully applied to molecular graph generation. Nevertheless, mitigating the computational complexity, which increases with the number of nodes in a graph, has been a major challenge. This has hindered the application of deep learning-based molecular graph generation to large molecules with many heavy atoms. In this study, we present a molecular graph compression method to alleviate the complexity while maintaining the capability of generating chemically valid and diverse molecular graphs. We designate six small substructural patterns that are prevalent between two atoms in real-world molecules. These relevant substructures in a molecular graph are then converted to edges by regarding them as additional edge features along with the bond types. This reduces the number of nodes significantly without any information loss. Consequently, a generative model can be constructed in a more efficient and scalable manner with large molecules on a compressed graph representation. We demonstrate the effectiveness of the proposed method for molecules with up to 88 heavy atoms using the GuacaMol benchmark.

Efficient learning of non-autoregressive graph variational autoencoders for molecular graph generation.

With the advancements in deep learning, deep generative models combined with graph neural networks have been successfully employed for data-driven molecular graph generation. Early methods based on the non-autoregressive approach have been effective in generating molecular graphs quickly and efficiently but have suffered from low performance. In this paper, we present an improved learning method involving a graph variational autoencoder for efficient molecular graph generation in a non-autoregressive manner. We introduce three additional learning objectives and incorporate them into the training of the model: approximate graph matching, reinforcement learning, and auxiliary property prediction. We demonstrate the effectiveness of the proposed method by evaluating it for molecular graph generation tasks using QM9 and ZINC datasets. The model generates molecular graphs with high chemical validity and diversity compared with existing non-autoregressive methods. It can also conditionally generate molecular graphs satisfying various target conditions.

Molecular Geometry Prediction using a Deep Generative Graph Neural Network.

A molecule's geometry, also known as conformation, is one of a molecule's most important properties, determining the reactions it participates in, the bonds it forms, and the interactions it has with other molecules. Conventional conformation generation methods minimize hand-designed molecular force field energy functions that are often not well correlated with the true energy function of a molecule observed in nature. They generate geometrically diverse sets of conformations, some of which are very similar to the lowest-energy conformations and others of which are very different. In this paper, we propose a conditional deep generative graph neural network that learns an energy function by directly learning to generate molecular conformations that are energetically favorable and more likely to be observed experimentally in data-driven manner. On three large-scale datasets containing small molecules, we show that our method generates a set of conformations that on average is far more likely to be close to the corresponding reference conformations than are those obtained from conventional force field methods. Our method maintains geometrical diversity by generating conformations that are not too similar to each other, and is also computationally faster. We also show that our method can be used to provide initial coordinates for conventional force field methods. On one of the evaluated datasets we show that this combination allows us to combine the best of both methods, yielding generated conformations that are on average close to reference conformations with some very similar to reference conformations.

Conditional Molecular Design with Deep Generative Models.

Although machine learning has been successfully used to propose novel molecules that satisfy desired properties, it is still challenging to explore a large chemical space efficiently. In this paper, we present a conditional molecular design method that facilitates generating new molecules with desired properties. The proposed model, which simultaneously performs both property prediction and molecule generation, is built as a semisupervised variational autoencoder trained on a set of existing molecules with only a partial annotation. We generate new molecules with desired properties by sampling from the generative distribution estimated by the model. We demonstrate the effectiveness of the proposed model by evaluating it on drug-like molecules. The model improves the performance of property prediction by exploiting unlabeled molecules and efficiently generates novel molecules fulfilling various target conditions.

Personalized prediction of drug efficacy for diabetes treatment via patient-level sequential modeling with neural networks.

Patients with type 2 diabetes mellitus are generally under continuous long-term medical treatment based on anti-diabetic drugs to achieve the desired glucose level. Thus, each patient is associated with a sequence of multiple records for prescriptions and their efficacies. Sequential dependencies are embedded in these records as personal factors so that previous records affect the efficacy of the current prescription for each patient. In this study, we present a patient-level sequential modeling approach utilizing the sequential dependencies to render a personalized prediction of the prescription efficacy. The prediction models are implemented using recurrent neural networks that use the sequence of all the previous records as inputs to predict the prescription efficacy at the time the current prescription is provided for each patient. Through this approach, each patient's historical records are effectively incorporated into the prediction. The experimental results of both the regression and classification analyses on real-world data demonstrate improved prediction accuracy, particularly for those patients having multiple previous records.

The usefulness of flexible cystoscopy for preventing double-J stent malposition after laparoscopic ureterolithotomy.

BACKGROUND: The aim of this study was to evaluate the role of flexible cystoscopy in preventing malpositioning of the ureteral stent after laparoscopic ureterolithotomy in male patients. METHODS: From April 2009 to June 2015, 97 male patients with stones >1.8 cm in the upper ureter underwent intracorporeal double-J stenting of the ureter after laparoscopic ureterolithotomy performed by four different surgeons. In the last 50 patients who underwent laparoscopic ureterolithotomy flexible cystoscopy was performed through the urethral route to confirm the position of the double-J stent, while in the first 47 correct positioning of the stent was confirmed through postoperative KUB. The demographic data and perioperative outcomes were reviewed retrospectively. Penalized logistic regression analysis was used to evaluate the effects of flexible cystoscopy. RESULTS: Upward malpositioning of the ureteral stent was found in 9 of the 47 (19.1%) patients who underwent surgery without flexible cystoscopy. Among the 50 most recent patients who underwent surgery with flexible cystoscopy through the urethral route, upward malpositioning was observed in 10 (20%) patients. The factors preventing upward malpositioning of the double-J catheter in multivariate analysis were surgeon (p = 0.039) and use of flexible cystoscopy (p = 0.008). CONCLUSION: Flexible cystoscopy is a simple, safe, quick, and effective method to identify and correct malpositioning of double-J stents, especially in male patients. TRIAL REGISTRATION: This study was registered with ClinicalTrials.gov Registry on May 11, 2017 (retrospective registration) with a trial registration number of NCT03150446 .

Effect of laminin 332 on motility and invasion in bladder cancer.

We examined the correlation between laminin 332 and malignancy in bladder cancer patients, and, using a strain of invasive bladder cancer cells, determined whether laminin 332 causes bladder cancer motility and invasion. To investigate the correlation between laminin 332 g2 distribution and patient outcome, we performed a semiquantitative immunohistochemical analysis of 35 paraffin-embedded samples using the antibody D4B5, which is specific for the laminin 5 γ2 chain. To evaluate the role of laminin 332 in NBT-II cell motility and invasion, we used a scratch assay and the Boyden chamber chemoinvasion system. Tumor stage and grade were significantly correlated with a loss of laminin 332 γ2 chain from the basement membrane (p = 0.001) and its retention in the cytoplasm (p = 0.001) (Kruskal-Wallis test). Kaplan-Meier survival curves revealed an association between the risk of progression and cytoplasmic retention of the laminin 332 γ2 chain. In addition, an in vitro scratch assay showed an increase in the migration of cells treated with laminin 332 from their cluster. The Boyden chamber assay showed that laminin 332 potentiated NBT-II cell invasion. Immunohistochemistry results showed that bladder cancer patients with a higher malignancy expressed more laminin 332. The in vitro scratch and invasion assay showed that laminin 332 stimulated the motility and invasion of bladder cancer cells. The invasion assay explains the correlation between laminin 332 expression and bladder cancer malignancy.

Simultaneous determination of nine UV filters and four preservatives in suncare products by high-performance liquid chromatography.

HPLC method for quantitative determination of four preservatives and nine UV filters worldwide authorized in commercial suncare product was developed and validated, and then 101 samples of commercial suncare products were analyzed for the UV filters and preservatives using the proposed method. The mobile phase was acetonitrile-water containing 0.5% acetic acid using a gradient elution at a flow rate of 0.9 mL/min and UV measurements were carried out at 320 nm for UV filters and 254 nm for preservatives. The correlation coefficients of each calibration curves were mostly higher than 0.999. The percent relative standard deviations (%RSD) ranged from 0.97% to 6.1% for five sample aliquots. The recoveries from the spiked solutions were 98-102%. 2-ethylhexyl-p-methoxycinnamate (EHMC) was detected in 96 of 101 commercial suncare products and the concentration was in the range of 3.08-8.16% and 18 samples were found to exceed the 7.5% which has been defined as the maximum allowed concentration in Korea. Methyl paraben was detected in 81 of 101 samples and the next-most often detected preservatives were propyl paraben (25), ethyl paraben (18), and butyl paraben (4). Three samples of 101 suncare products exceeded the maximum allowed concentration (i.e., 0.58-0.79%). The proposed HPLC method allows efficient and simultaneous analysis of preservatives and UV filters suitable for quality control assays of commercial suncare products.