Showing 1–6 of 6 results for author: Feinberg, E

Atomic Convolutional Networks for Predicting Protein-Ligand Binding Affinity

Authors: Joseph Gomes, Bharath Ramsundar, Evan N. Feinberg, Vijay S. Pande

Abstract: Empirical scoring functions based on either molecular force fields or cheminformatics descriptors are widely used, in conjunction with molecular docking, during the early stages of drug discovery to predict potency and binding affinity of a drug-like molecule to a given target. These models require expert-level knowledge of physical chemistry and biology to be encoded as hand-tuned parameters or f… ▽ More Empirical scoring functions based on either molecular force fields or cheminformatics descriptors are widely used, in conjunction with molecular docking, during the early stages of drug discovery to predict potency and binding affinity of a drug-like molecule to a given target. These models require expert-level knowledge of physical chemistry and biology to be encoded as hand-tuned parameters or features rather than allowing the underlying model to select features in a data-driven procedure. Here, we develop a general 3-dimensional spatial convolution operation for learning atomic-level chemical interactions directly from atomic coordinates and demonstrate its application to structure-based bioactivity prediction. The atomic convolutional neural network is trained to predict the experimentally determined binding affinity of a protein-ligand complex by direct calculation of the energy associated with the complex, protein, and ligand given the crystal structure of the binding pose. Non-covalent interactions present in the complex that are absent in the protein-ligand sub-structures are identified and the model learns the interaction strength associated with these features. We test our model by predicting the binding free energy of a subset of protein-ligand complexes found in the PDBBind dataset and compare with state-of-the-art cheminformatics and machine learning-based approaches. We find that all methods achieve experimental accuracy and that atomic convolutional networks either outperform or perform competitively with the cheminformatics based methods. Unlike all previous protein-ligand prediction systems, atomic convolutional networks are end-to-end and fully-differentiable. They represent a new data-driven, physics-based deep learning model paradigm that offers a strong foundation for future improvements in structure-based bioactivity prediction. △ Less

Submitted 30 March, 2017; originally announced March 2017.

Irradiation of Materials with Short, Intense Ion pulses at NDCX-II

Authors: P. A. Seidl, Q. Ji, A. Persaud, E. Feinberg, B. Ludewigt, M. Silverman, A. Sulyman, W. L. Waldron, T. Schenkel, J. J. Barnard, A. Friedman, D. P. Grote, E. P. Gilson, I. D. Kaganovich, A. D. Stepanov, F. Treffert, M. Zimmer

Abstract: We present an overview of the performance of the Neutralized Drift Compression Experiment-II (NDCX-II) accelerator at Berkeley Lab, and report on recent target experiments on beam driven melting and transmission ion energy loss measurements with nanosecond and millimeter-scale ion beam pulses and thin tin foils. Bunches with around 10^11 ions, 1-mm radius, and 2-30 ns FWHM duration have been creat… ▽ More We present an overview of the performance of the Neutralized Drift Compression Experiment-II (NDCX-II) accelerator at Berkeley Lab, and report on recent target experiments on beam driven melting and transmission ion energy loss measurements with nanosecond and millimeter-scale ion beam pulses and thin tin foils. Bunches with around 10^11 ions, 1-mm radius, and 2-30 ns FWHM duration have been created with corresponding fluences in the range of 0.1 to 0.7 J/cm^2. To achieve these short pulse durations and mm-scale focal spot radii, the 1.1 MeV He+ ion beam is neutralized in a drift compression section, which removes the space charge defocusing effect during final compression and focusing. The beam space charge and drift compression techniques resemble necessary beam conditions and manipulations in heavy ion inertial fusion accelerators. Quantitative comparison of detailed particle-in-cell simulations with the experiment play an important role in optimizing accelerator performance. △ Less

Submitted 12 April, 2017; v1 submitted 16 March, 2017; originally announced March 2017.

Comments: 15 pages, 7 figures. revised manuscript submitted to Laser and Particle Beams

MoleculeNet: A Benchmark for Molecular Machine Learning

Authors: Zhenqin Wu, Bharath Ramsundar, Evan N. Feinberg, Joseph Gomes, Caleb Geniesse, Aneesh S. Pappu, Karl Leswing, Vijay Pande

Abstract: Molecular machine learning has been maturing rapidly over the last few years. Improved methods and the presence of larger datasets have enabled machine learning algorithms to make increasingly accurate predictions about molecular properties. However, algorithmic progress has been limited due to the lack of a standard benchmark to compare the efficacy of proposed methods; most new algorithms are be… ▽ More Molecular machine learning has been maturing rapidly over the last few years. Improved methods and the presence of larger datasets have enabled machine learning algorithms to make increasingly accurate predictions about molecular properties. However, algorithmic progress has been limited due to the lack of a standard benchmark to compare the efficacy of proposed methods; most new algorithms are benchmarked on different datasets making it challenging to gauge the quality of proposed methods. This work introduces MoleculeNet, a large scale benchmark for molecular machine learning. MoleculeNet curates multiple public datasets, establishes metrics for evaluation, and offers high quality open-source implementations of multiple previously proposed molecular featurization and learning algorithms (released as part of the DeepChem open source library). MoleculeNet benchmarks demonstrate that learnable representations are powerful tools for molecular machine learning and broadly offer the best performance. However, this result comes with caveats. Learnable representations still struggle to deal with complex tasks under data scarcity and highly imbalanced classification. For quantum mechanical and biophysical datasets, the use of physics-aware featurizations can be more important than choice of particular learning algorithm. △ Less

Submitted 25 October, 2018; v1 submitted 1 March, 2017; originally announced March 2017.

Staging of RF-accelerating units in a MEMS-based ion accelerator

Authors: A. Persaud, P. A. Seidl, Q. Ji, E. Feinberg, W. L. Waldron, T. Schenkel, S. Ardanuc, K. B. Vinayakumar, A. Lal

Abstract: Multiple Electrostatic Quadrupole Array Linear Accelerators (MEQALACs) provide an opportunity to realize compact radio-frequency (RF) accelerator structures that can deliver very high beam currents. MEQALACs have been previously realized with acceleration gap distances and beam aperture sizes of the order of centimeters. Through advances in Micro-Electro-Mechanical Systems (MEMS) fabrication, MEQA… ▽ More Multiple Electrostatic Quadrupole Array Linear Accelerators (MEQALACs) provide an opportunity to realize compact radio-frequency (RF) accelerator structures that can deliver very high beam currents. MEQALACs have been previously realized with acceleration gap distances and beam aperture sizes of the order of centimeters. Through advances in Micro-Electro-Mechanical Systems (MEMS) fabrication, MEQALACs can now be scaled down to the sub-millimeter regime and batch processed on wafer substrates. In this paper, we show first results from using three RF stages in a compact MEMS-based ion accelerator. The results presented show proof-of-concept with accelerator structures formed from printed circuit boards using a 3x3 beamlet arrangement and noble gas ions at 10 keV. We present a simple model to describe the measured results. The model is then used to examine some of the aspects of this approach, such as possible effects of alignment errors. We also discuss some of the scaling behaviour of a compact MEQALAC. The MEMS-based approach enables a low-cost, highly versatile accelerator covering a wide range of beam energies and currents. Applications include ion-beam analysis, mass spectrometry, materials processing, and at very high beam powers, plasma heating. △ Less

Submitted 31 October, 2017; v1 submitted 1 February, 2017; originally announced February 2017.

Comments: Conference on the Application of Accelerators in Research and Industry, CAARI 2016, 30 October-4 November 2016, Ft. Worth, TX, USA

Journal ref: Physics Procedia, Volume 90, 2017, Pages 136-142

A compact linear accelerator based on a scalable microelectromechanical-system RF-structure

Authors: A. Persaud, Q. Ji, E. Feinberg, P. A. Seidl, W. L. Waldron, A. Lal, K. B. Vinayakumar, S. Ardanuc, D. A. Hammer, T. Schenkel

Abstract: A new approach for a compact radio-frequency (RF) accelerator structure is presented. The new accelerator architecture is based on the Multiple Electrostatic Quadrupole Array Linear Accelerator (MEQALAC) structure that was first developed in the 1980s. The MEQALAC utilized RF resonators producing the accelerating fields and providing for higher beam currents through parallel beamlets focused using… ▽ More A new approach for a compact radio-frequency (RF) accelerator structure is presented. The new accelerator architecture is based on the Multiple Electrostatic Quadrupole Array Linear Accelerator (MEQALAC) structure that was first developed in the 1980s. The MEQALAC utilized RF resonators producing the accelerating fields and providing for higher beam currents through parallel beamlets focused using arrays of electrostatic quadrupoles (ESQs). While the early work obtained ESQs with lateral dimensions on the order of a few centimeters, using printed circuits board (PCB), we reduce the characteristic dimension to the millimeter regime, while massively scaling up the potential number of parallel beamlets. Using Microelectromechanical systems scalable fabrication approaches, we are working on further reducing the characteristic dimension to the sub-millimeter regime. The technology is based on RF-acceleration components and ESQs implemented in PCB or silicon wafers where each beamlet passes through beam apertures in the wafer. The complete accelerator is then assembled by stacking these wafers. This approach has the potential for fast and inexpensive batch fabrication of the components and flexibility in system design for application specific beam energies and currents. For prototyping the accelerator architecture, the components have been fabricated using PCB. In this paper, we present proof of concept results of the principal components using PCB: RF acceleration and ESQ focusing. Ongoing developments on implementing components in silicon and scaling of the accelerator technology to high currents and beam energies are discussed. △ Less

Submitted 29 June, 2017; v1 submitted 30 October, 2016; originally announced October 2016.

Journal ref: Review of Scientific Instruments 88, 063304 (2017)

Recent Experiments At Ndcx-II: Irradiation Of Materials Using Short, Intense Ion Beams

Authors: P. A. Seidl, Q. Ji, A. Persaud, E. Feinberg, B. Ludewigt, M. Silverman, A. Sulyman, W. L. Waldron, T. Schenkel, J. J. Barnard, A. Friedman, D. P. Grote, E. P. Gilson, I. D. Kaganovich, A. Stepanov, F. Treffert, M. Zimmer

Abstract: We present an overview of the performance of the Neutralized Drift Compression Experiment-II (NDCX-II) accelerator at Berkeley Lab, and summarize recent studies of material properties created with nanosecond and millimeter-scale ion beam pulses. The scientific topics being explored include the dynamics of ion induced damage in materials, materials synthesis far from equilibrium, warm dense matter… ▽ More We present an overview of the performance of the Neutralized Drift Compression Experiment-II (NDCX-II) accelerator at Berkeley Lab, and summarize recent studies of material properties created with nanosecond and millimeter-scale ion beam pulses. The scientific topics being explored include the dynamics of ion induced damage in materials, materials synthesis far from equilibrium, warm dense matter and intense beam-plasma physics. We summarize the improved accelerator performance, diagnostics and results of beam-induced irradiation of thin samples of, e.g., tin and silicon. Bunches with over 3x10^10 ions, 1- mm radius, and 2-30 ns FWHM duration have been created. To achieve these short pulse durations and mm-scale focal spot radii, the 1.2 MeV He+ ion beam is neutralized in a drift compression section which removes the space charge defocusing effect during final compression and focusing. Quantitative comparison of detailed particle-in-cell simulations with the experiment play an important role in optimizing accelerator performance; these keep pace with the accelerator repetition rate of ~1/minute. △ Less

Submitted 17 October, 2016; originally announced October 2016.

Comments: Submitted to the 2016 North American Particle Accelerator Conference (NAPAC16). Proceedings to be published at http://jacow.org/