Machine learning (ML) and artificial intelligence (AI) methodologies are having a revolutionizing impact in advancing biomolecular imaging and informatics, improving medical diagnosis, and enabling drug discovery. In particular, ML approaches have impacted cryogenic electron microscopy (cryo-EM) and cryogenic electron tomography (cryo-ET) workflows in areas spanning data collection to reconstruction and model building. These approaches are proving effective in automating labor-intensive tasks and in extracting information from the massive, but low-signal datasets that microscopy typically produces. Additionally, careful application of those methods offers a chance to naturally encode prior structural information and orthogonal biochemical and biophysical studies into the information analysis pipeline. Applying these approaches to biomedical imaging techniques and structural science studies has accelerated research outcomes and opened new avenues of investigation.
ML and AI is transforming cryo-EM by accelerating workflows, and enriching data analysis, processing, and annotation. Rapid advances in hardware and image processing algorithms have aided ML adoption by researchers to better understand and predict the structures of biological molecules and relate them to their functional properties. ML has powered many aspects of cryo-EM structure determination and greatly promoted its development by incorporating biophysical knowledge into our algorithms. When misapplied, however, the chances of introducing difficult-to-identify artifacts increases, and necessitates the establishment of robust validation methods and best practices. To promote growth in ML methods and adoption of their applications to electron microscopy there is an increased movement to establish databases of annotated training and test sets. As the theoretical and experimental approaches converge the field is overcoming challenges towards integration into every step of the structural biology workflow and having a transformational impact on biomedical sciences.
This Research Topic invites commentaries, reviews, perspectives, or technology and methods papers using deep learning to improve the determination of structures in biological macromolecules in cryogenic electron microscopy. Examples include, but are not limited to:
• Highlighting the strengths, future prospects, and potential concerns of ML and AI approaches in cryo-EM, as well as in combined approaches in protein engineering and other modalities of structural biology.
• Novel approaches and algorithms for accelerating structure determination and image analysis to methods for modeling protein sequence, structure, function, and beyond.
• FAIR (Findable, Accessible, Interoperable, and Reusable) use of data and information sharing to aid establishment of validation metrics and best practices.
• Integrated computational and experimental studies where AI methods support experimental validation.*
*Articles submitted to this collection will however be encouraged to incorporate experimental data used to validate the ensued in silico predictions.
Ruben Sanchez Garcia is a post-doctoral fellow funded via an Astex Pharmaceuticals Sustaining Innovation Post-Doctoral Fellowship. All other Topic Editors declare no competing interests.
Machine learning (ML) and artificial intelligence (AI) methodologies are having a revolutionizing impact in advancing biomolecular imaging and informatics, improving medical diagnosis, and enabling drug discovery. In particular, ML approaches have impacted cryogenic electron microscopy (cryo-EM) and cryogenic electron tomography (cryo-ET) workflows in areas spanning data collection to reconstruction and model building. These approaches are proving effective in automating labor-intensive tasks and in extracting information from the massive, but low-signal datasets that microscopy typically produces. Additionally, careful application of those methods offers a chance to naturally encode prior structural information and orthogonal biochemical and biophysical studies into the information analysis pipeline. Applying these approaches to biomedical imaging techniques and structural science studies has accelerated research outcomes and opened new avenues of investigation.
ML and AI is transforming cryo-EM by accelerating workflows, and enriching data analysis, processing, and annotation. Rapid advances in hardware and image processing algorithms have aided ML adoption by researchers to better understand and predict the structures of biological molecules and relate them to their functional properties. ML has powered many aspects of cryo-EM structure determination and greatly promoted its development by incorporating biophysical knowledge into our algorithms. When misapplied, however, the chances of introducing difficult-to-identify artifacts increases, and necessitates the establishment of robust validation methods and best practices. To promote growth in ML methods and adoption of their applications to electron microscopy there is an increased movement to establish databases of annotated training and test sets. As the theoretical and experimental approaches converge the field is overcoming challenges towards integration into every step of the structural biology workflow and having a transformational impact on biomedical sciences.
This Research Topic invites commentaries, reviews, perspectives, or technology and methods papers using deep learning to improve the determination of structures in biological macromolecules in cryogenic electron microscopy. Examples include, but are not limited to:
• Highlighting the strengths, future prospects, and potential concerns of ML and AI approaches in cryo-EM, as well as in combined approaches in protein engineering and other modalities of structural biology.
• Novel approaches and algorithms for accelerating structure determination and image analysis to methods for modeling protein sequence, structure, function, and beyond.
• FAIR (Findable, Accessible, Interoperable, and Reusable) use of data and information sharing to aid establishment of validation metrics and best practices.
• Integrated computational and experimental studies where AI methods support experimental validation.*
*Articles submitted to this collection will however be encouraged to incorporate experimental data used to validate the ensued in silico predictions.
Ruben Sanchez Garcia is a post-doctoral fellow funded via an Astex Pharmaceuticals Sustaining Innovation Post-Doctoral Fellowship. All other Topic Editors declare no competing interests.