Applied AI letters最新文献

{"title":"Assessing methods and obstacles in chemical space exploration","authors":"Shawn Reeves,&nbsp;Benjamin DiFrancesco,&nbsp;Vijay Shahani,&nbsp;Stephen MacKinnon,&nbsp;Andreas Windemuth,&nbsp;Andrew E. Brereton","doi":"10.1002/ail2.17","DOIUrl":"https://doi.org/10.1002/ail2.17","url":null,"abstract":"<p>Benchmarking the performance of generative methods for drug design is complex and multifaceted. In this report, we propose a separation of concerns for de novo drug design, categorizing the task into three main categories: <i>generation</i>, <i>discrimination</i> and <i>exploration</i>. We demonstrate that changes to any of these three concerns impacts benchmark performance for drug design tasks. In this report, we present Deriver, an open-source Python package that acts as a modular framework for molecule generation, with a focus on integrating multiple generative methods. Using Deriver, we demonstrate that changing parameters related to each of these three concerns impacts chemical space traversal significantly, and that the freedom to independently adjust each is critical to real-world applications having conflicting priorities. We find that combining multiple generative methods can improve optimization of molecular properties and lower the chance of becoming trapped in local minima. In addition, filtering molecules for drug-likeness (based on physicochemical properties and SMARTS pattern matching) before they are scored may hinder exploration, but can also improve the quality of the final molecules. Finally, we demonstrate that any given task has an exploration algorithm best suited to it, though in practice linear probabilistic sampling generally results in the best outcomes, when compared to Monte Carlo sampling or greedy sampling. Deriver is being made freely available, to help others interested in collaboratively improving existing methods in de novo drug design centered around inheritance of molecular structure, modularity, extensibility, and separation of concerns.</p>","PeriodicalId":72253,"journal":{"name":"Applied AI letters","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2020-11-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://sci-hub-pdf.com/10.1002/ail2.17","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"137848389","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Visualizing and understanding inherent features in SD-OCT for the progression of age-related macular degeneration using deconvolutional neural networks","authors":"Sajib Saha,&nbsp;Ziyuan Wang,&nbsp;Srinivas Sadda,&nbsp;Yogesan Kanagasingam,&nbsp;Zhihong Hu","doi":"10.1002/ail2.16","DOIUrl":"10.1002/ail2.16","url":null,"abstract":"<p>To develop a convolutional neural network visualization strategy so that optical coherence tomography (OCT) features contributing to the evolution of age-related macular degeneration (AMD) can be better determined. We have trained a U-Net model to utilize baseline OCT to predict the progression of geographic atrophy (GA), a late stage manifestation of AMD. We have augmented the U-Net architecture by attaching deconvolutional neural networks (deconvnets). Deconvnets produce the reconstructed feature maps and provide an indication regarding the inherent baseline OCT features contributing to GA progression. Experiments were conducted on longitudinal spectral domain (SD)-OCT and fundus autofluorescence images collected from 70 eyes with GA. The intensity of Bruch's membrane-outer choroid (BMChoroid) retinal junction exhibited a relative importance of 24%, in the GA progression. The intensity of the inner retinal pigment epithelium (RPE) and BM junction (InRPEBM) showed a relative importance of 22%. BMChoroid (where the AMD feature/damage of choriocapillaris was included) followed by InRPEBM (where the AMD feature/damage of RPE was included) are the layers which appear to be most relevant in predicting the progression of AMD.</p>","PeriodicalId":72253,"journal":{"name":"Applied AI letters","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2020-10-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://sci-hub-pdf.com/10.1002/ail2.16","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"10372028","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Multiobjective genetic programming for reinforced concrete beam modeling","authors":"Amirhessam Tahmassebi,&nbsp;Behshad Mohebali,&nbsp;Anke Meyer-Baese,&nbsp;Amir H. Gandomi","doi":"10.1002/ail2.9","DOIUrl":"10.1002/ail2.9","url":null,"abstract":"<p>This paper presents the application of multiobjective genetic programming (MOGP) in engineering issues. An evolutionary symbolic implementation was developed based on a case study on prediction of the shear strength of slender reinforced concrete beams without stirrups including 1942 set of published test results. In the implementation of the MOGP model, the nondominated sorting genetic algorithm II with adaptive regression by mixing algorithm with considering the optimization of mean-square error as the fitness measure and the subtree complexity was used. The developed MOGP model was compared to previously developed genetic programming models, different building codes, and additional machine learning based approaches. It is clearly shown that the MOGP model outperformed the other algorithms applied on this database and can be a general solution on any engineering problems with the main advantage of prediction equations without assuming prior form of the relevance among the input predictor variables.</p>","PeriodicalId":72253,"journal":{"name":"Applied AI letters","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2020-09-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://sci-hub-pdf.com/10.1002/ail2.9","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"49213040","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Practical notes on building molecular graph generative models","authors":"Rocío Mercado, T. Rastemo, Edvard Lindelöf, G. Klambauer, O. Engkvist, Hongming Chen, E. Bjerrum","doi":"10.26434/chemrxiv.12888383","DOIUrl":"https://doi.org/10.26434/chemrxiv.12888383","url":null,"abstract":"Here are presented technical notes\u0000and tips on developing graph generative models for molecular design. This work\u0000stems from the development of GraphINVENT, a Python platform for graph-based molecular\u0000generation using graph neural networks. In this work, technical details that\u0000could be of interest to researchers developing their own molecular generative\u0000models are discussed, including strategies for designing new models. Advice on development and debugging tools\u0000which were helpful during code development is also provided. Finally, methods that were tested but which ultimately\u0000didn’t lead to promising results in the development of GraphINVENT are\u0000described here in the hope that this will help other researchers avoid pitfalls\u0000in development and instead focus their efforts on more promising strategies for\u0000graph-based molecular generation.","PeriodicalId":72253,"journal":{"name":"Applied AI letters","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2020-08-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"44946469","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Assessing methods and obstacles in chemical space exploration","authors":"Shawn Reeves, B. Difrancesco, V. Shahani, S. MacKinnon, A. Windemuth, Andrew E. Brereton","doi":"10.26434/chemrxiv.12761840","DOIUrl":"https://doi.org/10.26434/chemrxiv.12761840","url":null,"abstract":"Benchmarking the performance of generative methods for drug design is complex and multifaceted. In this report, we propose a separation of concerns for de novo drug design, categorizing the task into three main categories: generation, discrimination, and exploration. We demonstrate that changes to any of these three concerns impacts benchmark performance for drug design tasks. In this report we present Deriver, an open-source Python package that acts as a modular framework for molecule generation, with a focus on integrating multiple generative methods. Using Deriver, we demonstrate that changing parameters related to each of these three concerns impacts chemical space traversal significantly, and that the freedom to independently adjust each is critical to real-world applications having conflicting priorities. We find that combining multiple generative methods can improve optimization of molecular properties, and lower the chance of becoming trapped in local minima. Additionally, filtering molecules for drug-likeness (based on physicochemical properties and SMARTS pattern matching) before they are scored can hinder exploration, but can improve the quality of the final molecules. Finally, we demonstrate that any given task has an exploration algorithm best suited to it, though in practice linear probabilistic sampling generally results in the best outcomes, when compared to Monte Carlo sampling or greedy sampling. We intend that Deriver, which is being made freely available, will be helpful to others interested in collaboratively improving existing methods in de novo drug design centered around inheritance of molecular structure, modularity, extensibility, and separation of concerns.","PeriodicalId":72253,"journal":{"name":"Applied AI letters","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2020-08-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"48509486","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Welcome to the first issue of Applied AI Letters","authors":"Edward O. Pyzer-Knapp,&nbsp;James Cuff,&nbsp;Jack Patterson,&nbsp;Olexandr Isayev,&nbsp;Simon Maskell","doi":"10.1002/ail2.8","DOIUrl":"10.1002/ail2.8","url":null,"abstract":"<p>It is a pleasure and an honour to welcome you to the first edition of <i>Applied AI Letters</i>. Getting to this point has been a combination of many people's hard work and we are very excited to move into the next stage, sharing our vision for <i>Applied AI Letters</i> with you.</p><p>When we consider the lifecycle of a successful idea, we can identify some unique stages. We have put these together in Figure 1. Initially, a challenge should be identified, and it is often (although not always) the case that there is some idea of the impact solving this challenge should have. If this challenge ignites the scientific spirit, research then begins <i>post haste</i>. Hopefully, at some point a research breakthrough is made, allowing a first in kind or first in class solution to the challenge. Solving a challenge is not the end of the story though, with the approach then being generalised, allowing its application to new and more impactful areas. Finally, the idea becomes so ingrained in the scientific and technological psyche that its use is no longer questioned, but expected; we call this the commoditisation stage. This way of thinking has been adopted in many places, such as the technology adoption curve, and the TRL levels used by, for example, NASA and the EU.<span>¹</span></p><p>Research progresses from a challenge being set down through the initial breakthrough, to the generalisation of the idea which enables its widespread application in new and novel areas and finally its commoditisation.</p><p>At <i>Applied AI Letters</i>, we noted that, in the field of AI, whilst there is an abundance of places that one can publicise the “breakthrough” stage, it was much harder to find places to communicate impactful applications and innovations. Additionally, we felt that there was an abundance of potential for impact if we could start a venue in which people across disciplines could communicate their solutions to challenges posed in a wide variety of potential application areas.</p><p>We strongly believe that we improve the sustainability of long-term AI research through communicating and celebrating the evolution from idea to real world application. In order to achieve this, there are many additional considerations beyond the theoretical and methodological elements of successfully deploying and delivering an AI application in the real world, and we need a forum to discuss and communicate these successes with new potential users.</p><p>AI has the potential to revolutionise many areas of research. At <i>Applied AI Letters</i>, we aim to be a world-leading venue for the demonstration of, and dissusion about, the application of cutting edge AI technologies to the most impactful problems of today. We belive that openness, fairness and diversity are key to delivering on this promise, and want to foster a culture which embraces these key principles. With applied AI evolving at such a rapid rate, we appreciate that speed is important for the timely ","PeriodicalId":72253,"journal":{"name":"Applied AI letters","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2020-08-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://sci-hub-pdf.com/10.1002/ail2.8","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"111069566","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}