In the space of only a few years, deep generative modeling has revolutionized how we think of artificial creativity, yielding autonomous systems which produce original images, music, and text. Inspired by these successes, researchers are now applying deep generative modeling techniques to the generation and optimization of molecules-in our review we found 45 papers on the subject published in the past two years. These works point to a future where such systems will be used to generate lead molecules, greatly reducing resources spent downstream synthesizing and characterizing bad leads in the lab. In this review we survey the increasingly complex landscape of models and representation schemes that have been proposed. The four classes of techniques we describe are recursive neural networks, autoencoders, generative adversarial networks, and reinforcement learning. After first discussing some of the mathematical fundamentals of each technique, we draw high level connections and comparisons with other techniques and expose the pros and cons of each. Several important high level themes emerge as a result of this work, including the shift away from the SMILES string representation of molecules towards more sophisticated representations such as graph grammars and 3D representations, the importance of reward function design, the need for better standards for benchmarking and testing, and the benefits of adversarial training and reinforcement learning over maximum likelihood based training.
We present a proof of concept that machine learning techniques can be used to predict the properties of CNOHF energetic molecules from their molecular structures. We focus on a small but diverse dataset consisting of 109 molecular structures spread across ten compound classes. Up until now, candidate molecules for energetic materials have been screened using predictions from expensive quantum simulations and thermochemical codes. We present a comprehensive comparison of machine learning models and several molecular featurization methods - sum over bonds, custom descriptors, Coulomb matrices, Bag of Bonds, and fingerprints. The best featurization was sum over bonds (bond counting), and the best model was kernel ridge regression. Despite having a small data set, we obtain acceptable errors and Pearson correlations for the prediction of detonation pressure, detonation velocity, explosive energy, heat of formation, density, and other properties out of sample. By including another dataset with ≈300 additional molecules in our training we show how the error can be pushed lower, although the convergence with number of molecules is slow. Our work paves the way for future applications of machine learning in this domain, including automated lead generation and interpreting machine learning models to obtain novel chemical insights.
Sources of support: This work was funded by the American Society for Radiation Oncology. Disclosures: All task force members' disclosure statements were reviewed before being invited and were shared with other task force members throughout the guideline's development. Those disclosures are published within this report. Where potential conflicts were detected, remedial measures to address them were taken.
11001 Background: The predominant pattern of failure of retroperitoneal sarcoma (RPS), frequently associated with subsequent death, is locoregional recurrence. Unlike in limbs, the efficacy of radiotherapy (RT) combined with surgery is not established. Methods: STRASS is a randomized, multicentre, international trial. Eligible patients had histologically-proven localized primary RPS, operable and suitable for radiotherapy. Patients were randomized 1:1 to preoperative RT (3D-CRT or IMRT) 50.4 Gy followed by surgery (RT/S group) or surgery alone (S group), stratified by hospital and performance status (0-1 vs 2). Primary endpoint is abdominal recurrence-free survival (ARFS; local relapse after complete resection, peritoneal sarcomatosis, R2 surgery, progressive disease during RT or unresectable disease). IDMC recommended a sensitivity analysis in which local progression on RT is not regarded as an event for patients who subsequently achieve complete surgical resection. Secondary endpoints were recurrence-free survival, overall survival, acute toxicity profile of RT, perioperative and late complications, and QoL. The study was designed to provide 90% power to show an increase of 20% in the 5-year ARFS rate, from 50% to 70% (corresponding to a HR of 0.52) at 2-sided 5% significance level. Results: 266 patients from Europe, USA and Canada were randomized between January 2012 and April 2017; 198 patients (74.5 %) had liposarcoma (LPS). Eighteen patients were designated ineligible. Overall rate of re-operation for any complication was 10.1%: 13 (10.9%) and 12 (9.4%) patients in RT/S versus S groups. 19 pts (14%) progressed during RT, 4 of whom did not undergo surgery. 3-year ARFS was 60.4% (95% Confidence interval (CI) 51.4-68.2%) and 58.7% (49.5- 66.7%) (HR = 1.01, 95%CI 0.71-1.44, p=0.954) in RT/S versus S groups. In the sensitivity analysis, 3-year ARFS was 66.0% (57.1-73.5%) and 58.7% (49.5-66.7%) in RT/S versus S groups (HR = 0.84, 95% CI 0.58-1.21, p=0.340). In the LPS subgroup, 3-year ARFS (sensitivity analysis) was 71.6% (61.3-79.6%) and 60.4% (49.8-69.5%) in RT/S versus S groups (HR = 0.64, 95%CI 0.40-1.01, p =0.049). Conclusion: STRASS failed to demonstrate a benefit of pre-operative RT for RPS. In the exploratory analysis, preoperative RT may benefit the LPS subgroup. Funding Source: EORTC and EUROSARC FP7 278472. Clinical trial information: EORTC 62092.
Water-graphene wetting interactions are central to several applications such as desalination, water filtration, electricity generation, biochemical sensing, fabrication of fuel cells, and many more. While substantial attention has been devoted to probe the wetting statics of a water drop on graphene, unraveling the possible wetting translucency nature of graphene, very little research has been done on the dynamics of wetting of water drops on graphene-coated solids or free-standing graphene layers. In this paper, we employ molecular dynamics (MD) simulations to study the contact and the spreading of a water nanodrop, quantifying its wetting dynamics, on supported and free-standing graphene. We demonstrate that nanoscale water drops establish contact with graphene by forming patches on graphene, and this patch formation is hastened for graphene layer(s) supported on hydrophilic solids. More importantly, our results demonstrate that the nanodrop spreading dynamics, regardless of the number of graphene layers or the nature of the underlying solid, obey the half-power law, i.e., r∼t(1/2) (where r is the wetting contact radius and t is the spreading time) for the entire timespan of spreading except towards the very end of the spreading lifetime when the spreading stops. Such a spreading behavior is exactly analogous to the spreading dynamics of nanodroplets for standard solids - this is in sharp contrast to the wetting statics of graphene where the wetting translucency effect makes graphene different from other standard solids.
Prior analyses of transplant outcomes in lupus transplant recipients have not consisted of multivariate analyses in the modern immunosuppressive era. Here, we compared patient and graft outcomes in lupus and non-lupus recipients transplanted between 1996 to 2000 using the United Network of Organ Sharing/Organ Procurement Transplant Network database. We evaluated the impact of recipient and donor demographic factors, time on dialysis and the initial immunosuppression regimen on rejection rates and transplant outcomes. Univariate analysis showed similar graft but better patient survival rates for primary lupus and non-lupus transplant recipients (5-year patient survival rates for lupus cohort 85.2% for deceased donor transplants and 92.1% for living donor transplants as opposed to 82.1% and 89.8% respectively for the non-lupus cohort; P=0.05 and 0.03) but similar patient survival rates for deceased donor retransplant patients. After controlling for confounding factors, no differences in patient or graft survival were seen between the two groups. No difference in acute rejection rates were observed in deceased donor transplants, but there was a small but significant increase in the risk of acute rejection in living donor lupus transplant recipients (hazard ratio=1.19, P=0.05). Risk of graft failure was lower for deceased donor recipients receiving MMF (five-year graft loss rate=29.6% for MMF vs. 40.2% for those not receiving MMF, P<0.0001), but no differences were seen among living donor recipients. Outcomes were similar regardless of type of calcineurin inhibitor, induction therapy, and time on dialysis. We conclude that lupus transplant recipients have outcomes generally equivalent to non-lupus transplant recipients.
An all-atom MD framework is developed to investigate densely grafted polyelectrolyte (PE) brushes. The solvation water of counterions is replaced by charged functional groups on the PE chains. The complex between counterions and the negatively charged PE segments overwhelms water by weight and volume above a critical grafting density, giving rise to a ''water-in-salt''-like scenario. Furthermore, the counterions and water molecules lose most of their mobility within the PE brushes as a result of electrostatic interactions and brush-induced confinement.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.