Classification of Protein Crystallization Images using EfficientNet with Data Augmentation

Edwards, David; Dinc, Imren

doi:10.1145/3429210.3429220

Cited by 5 publications

(7 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…With the advent of convolutional neural networks (CNN), the MARCO model was trained and tested using a data set of 500,000 macromolecular crystal images, achieving an accurate identification rate of 94% . The accuracy of crystal recognition was further improved to 97.23% using a scalable deep CNN, named EfficientNet . Another CNN-based model called CrystalNet was developed, enabling the classification of protein crystals at different stages in time resolution during high-throughput screening, thereby facilitating the monitoring of crystal evolution .…”

Section: Screening and Optimization Of Crystallization Conditionsmentioning

confidence: 99%

Artificial Intelligence Assisted Pharmaceutical Crystallization

Zhu,

Zhang,

Wang

et al. 2024

Crystal Growth & Design

View full text Add to dashboard Cite

The ever-increasing demand for novel drug development has spurred the adaptation of conventional research methods in the era of artificial intelligence. Pharmaceutical crystallization, as an essential part of drug development, has become a thrilling research frontier that reduces screening labor via integrating automated high-throughput platforms with in situ monitoring and datadriven algorithms (e.g., machine learning) to predict physicochemical properties and various solid-state forms. In this review, we started with a primer to introduce the machine learning algorithms that are widely used in pharmaceutical crystallization. Then, we systematically summarized recent advancements on high-throughput platforms to acquire huge amounts of data sets, prediction of physicochemical properties based on abundant experimental data, optimization and monitoring of pharmaceutical crystallization process to screen crystallization conditions, and prediction of polymorphs and cocrystals. Finally, we discussed the challenges and opportunities in an endeavor to develop a fully automated pharmaceutical crystallization screening paradigm for ultimately realizing a selfdriving screening laboratory. This review highlights the frontier of artificial intelligence in pharmaceutical crystallization and offers a guideline for beginners to not only understand the basic principles of machine learning algorithms but also learn how to utilize machine learning to accelerate pharmaceutical crystallization development.

show abstract

Section: Screening and Optimization Of Crystallization Conditionsmentioning

confidence: 99%

Artificial Intelligence Assisted Pharmaceutical Crystallization

Zhu,

Zhang,

Wang

et al. 2024

Crystal Growth & Design

View full text Add to dashboard Cite

show abstract

“…Logo, o objetivo de aplicar estas transformac ¸ões nas imagens de entrada da rede é deixar que o algorítmo infira tal invariância, criando, assim, uma "noc ¸ão" do espac ¸o visual, o que resulta no aumento do potencial de generalizac ¸ão da rede [16,17]. Frequentemente são relatados bons resultados com o uso desta técnica [18][19][20], principalmente quando existe grande similaridade entre as classes.…”

Section: Aumento Artificial De Dadosunclassified

“…Em vez de dimensionar cada aspecto manualmente, o modelo implementa um escalonamento composto que equilibra os aspectos para obter melhor desempenho, com isso a rede consegue uma alta acurácia usando muito menos parâmtros e operac ¸ões de ponto flutuante por segundo (FLOPS). Esta rede já foi usada na classificac ¸ão de doenc ¸as em vegetais [33], eletrocardiogramas [18] e cristalizac ¸ão de proteínas [19].…”

Section: Efficientnetunclassified

Classificação Morfológica de Galáxias no S-PLUS por Combinação de Redes Convolucionais

Cardoso¹,

Schwarz²,

Dias³

et al. 2021

View full text Add to dashboard Cite

The universe is composed of galaxies that have diverse shapes. Once the structure of a galaxy is determined, it is possible to obtain important information about its formation and evolution. Morphologically classifying galaxies means cataloging them according to their visual appearance and the classification is linked to the physical properties of the galaxy. A morphological classification made through visual inspection is subject to biases introduced by subjective observations made by human volunteers. For this reason, systematic, objective and easily reproducible classification of galaxies has been gaining importance since the astronomer Edwin Hubble created his famous classification method. In this work, we combine accurate visual classifications of the Galaxy Zoo project with Deep Learning methods. The goal is to find an efficient technique at human performance level classification, but in a systematic and automatic way, for classification of elliptical and spiral galaxies. For this, a neural network model was created through an Ensemble of four other convolutional models, allowing a greater accuracy in the classification than what would be obtained with any one individual. Details of the individual models and improvements made are also described. The present work is entirely based on the analysis of images (not parameter tables) from DR1 (www.datalab.noao.edu) of the Southern Photometric Local Universe Survey (S-PLUS). In terms of classification, we achieved, with the Ensemble, an accuracy of ≈ 99% in the test sample (using pre-trained networks).

show abstract

“…Any ML application is dependent on the quality of the data used to train the system-the MARCO model used half a million images from 5 different groups, and gives impressive results [24]. However, during the process of training the MARCO model it was noticed that the model's performance on images from our lab (The Collaborative Crystallisation Centre…”

Section: Introductionmentioning

confidence: 99%

Moving beyond MARCO

Rosa

Watkins²,

Newman³

2023

PLoS ONE

View full text Add to dashboard Cite

The use of imaging systems in protein crystallisation means that the experimental setups no longer require manual inspection to determine the outcome of the trials. However, it leads to the problem of how best to find images which contain useful information about the crystallisation experiments. The adoption of a deeplearning approach in 2018 enabled a four-class machine classification system of the images to exceed human accuracy for the first time. Underpinning this was the creation of a labelled training set which came from a consortium of several different laboratories. The MARCO classification model does not have the same accuracy on local data as it does on images from the original test set; this can be somewhat mitigated by retraining the ML model and including local images. We have characterized the image data used in the original MARCO model, and performed extensive experiments to identify training settings most likely to enhance the local performance of a MARCO-dataset based ML classification model.

show abstract

Classification of Protein Crystallization Images using EfficientNet with Data Augmentation

Cited by 5 publications

References 18 publications

Artificial Intelligence Assisted Pharmaceutical Crystallization

Artificial Intelligence Assisted Pharmaceutical Crystallization

Classificação Morfológica de Galáxias no S-PLUS por Combinação de Redes Convolucionais

Moving beyond MARCO

Contact Info

Product

Resources

About