A massively parallel infrastructure for adaptive multiscale simulations

Natale, Francesco Di; Bhatia, Harsh; Carpenter, Timothy S.; Neale, Chris; Schumacher, Sara I. L. Kokkila; Oppelstrup, Tomas; Stanton, Liam; Zhang, Xiaohua; Sundram, Shiv; Scogland, Thomas R. W.; Dharuman, Gautham; Surh, Michael P.; Yang, Yue; Misale, Claudia; Schneidenbach, Lars; Costa, Carlos H. Ã.; Kim, Changhoan; D'Amora, Bruce; Gnanakaran, S.; Nissley, Dwight V.; Streitz, Fred; Lightstone, Felice C.; Bremer, Peer-Timo; Glosli, James N.; Ingólfsson, Helgi I.

doi:10.1145/3295500.3356197

Cited by 44 publications

(37 citation statements)

References 59 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…While only accounting for ∼2% of the inner leaflet, PIPs are known to be key lipids for interacting with certain proteins, and system A-8 was designed with that in mind. 59 Conversely, by limiting ourselves to just eight lipid types and deciding to include a glycolipid species in B-8 (as there are >10% glycolipids in the B-58 outer leaflet), it means compromising elsewhere by only having a single type of PE lipid. What is gained in terms of richness of headgroup diversity is relinquished in range of lipid tails.…”

Section: Discussionmentioning

confidence: 99%

Capturing Biologically Complex Tissue-Specific Membranes at Different Levels of Compositional Complexity

et al. 2020

Self Cite

View full text Add to dashboard Cite

Plasma membranes (PMs) contain hundreds of different lipid species that contribute differently to overall bilayer properties. By modulation of these properties, membrane protein function can be affected. Furthermore, inhomogeneous lipid mixing and domains of lipid enrichment/depletion can sort proteins and provide optimal local environments. Recent coarse-grained (CG) Martini molecular dynamics efforts have provided glimpses into lipid organization of different PMs: an “Average” and a “Brain” PM. Their high complexity and large size require long simulations (∼80 μs) for proper sampling. Thus, these simulations are computationally taxing. This level of complexity is beyond the possibilities of all-atom simulations, raising the question—what complexity is needed for “realistic” bilayer properties? We constructed CG Martini PM models of varying complexity (63 down to 8 different lipids). Lipid tail saturations and headgroup combinations were kept as consistent as possible for the “tissues’” (Average/Brain) at three levels of compositional complexity. For each system, we analyzed membrane properties to evaluate which features can be retained at lower complexity and validate eight-component bilayers that can act as reliable mimetics for Average or Brain PMs. Systems of reduced complexity deliver a more robust and malleable tool for computational membrane studies and allow for equivalent all-atom simulations and experiments.

show abstract

Section: Discussionmentioning

confidence: 99%

Capturing Biologically Complex Tissue-Specific Membranes at Different Levels of Compositional Complexity

et al. 2020

Self Cite

View full text Add to dashboard Cite

show abstract

“…However, the intended purpose of BD-GFRD is not to simulate biomolecular interactions with detailed biomolecular interactions. Specialized simulations (65) are in development to answer these questions, which can accommodate detailed biomolecular interactions but fail to reach experimentally accessible timescales. BD-GFRD takes a complementary approach by trading biological details for long timescale simulation.…”

Section: Discussionmentioning

confidence: 99%

Presence or Absence of Ras Dimerization Shows Distinct Kinetic Signature in Ras-Raf Interaction

Sarkar

Garcı́a

2020

Biophysical Journal

View full text Add to dashboard Cite

Initiations of cell signaling pathways often occur through the formation of multiprotein complexes that form through protein-protein interactions. Therefore, detecting their presence is central to understanding the function of a cell signaling pathway, aberration of which often leads to fatal diseases, including cancers. However, the multiprotein complexes are often difficult to detect using microscopes due to their small sizes. Therefore, currently, their presence can be only detected through indirect means. In this article, we propose to investigate the presence or absence of protein complexes through some easily measurable kinetic parameters, such as activation rates. As a proof of concept, we investigate the Ras-Raf system, a wellcharacterized cell signaling system. It has been hypothesized that Ras dimerization is necessary to create activated Raf dimers. Although there are circumstantial evidences supporting the Ras dimerization hypothesis, direct proof of Ras dimerization is still inconclusive. In the absence of conclusive direct experimental proof, this hypothesis can only be examined through indirect evidences of Ras dimerization. In this article, using a multiscale simulation technique, we provide multiple criteria that distinguishes an activation mechanism involving Ras dimerization from another mechanism that does not involve Ras dimerization. The provided criteria will be useful in the investigation of not only Ras-Raf interaction but also other two-protein interactions.

show abstract

“…To understand how our approach performs in loading verylarge datasets, we used another 892 GB dataset generated from molecular dynamics (MD) simulations conducted using Multiscale Machine-Learned Modeling Infrastructure (MuMMI) [23]. The dataset contains~7M files that are derived MD trajectory frames.…”

Section: Methodsmentioning

confidence: 99%

Accelerating Data Loading in Deep Neural Network Training

Yang

Cong

2019

2019 IEEE 26th International Conference on High Performance Computing, Data, and Analytics (HiPC)

View full text Add to dashboard Cite

Data loading can dominate deep neural network training time on large-scale systems. We present a comprehensive study on accelerating data loading performance in large-scale distributed training. We first identify performance and scalability issues in current data loading implementations. We then propose optimizations that utilize CPU resources to the data loader design. We use an analytical model to characterize the impact of data loading on the overall training time and establish the performance trend as we scale up distributed training. Our model suggests that I/O rate limits the scalability of distributed training, which inspires us to design a locality-aware data loading method. By utilizing software caches, our method can drastically reduce the data loading communication volume in comparison with the original data loading implementation. Finally, we evaluate the proposed optimizations with various experiments. We achieved more than 30x speedup in data loading using 256 nodes with 1,024 learners.

show abstract

A massively parallel infrastructure for adaptive multiscale simulations

Cited by 44 publications

References 59 publications

Capturing Biologically Complex Tissue-Specific Membranes at Different Levels of Compositional Complexity

Capturing Biologically Complex Tissue-Specific Membranes at Different Levels of Compositional Complexity

Presence or Absence of Ras Dimerization Shows Distinct Kinetic Signature in Ras-Raf Interaction

Accelerating Data Loading in Deep Neural Network Training

Contact Info

Product

Resources

About