The LoBoS cluster has been expanded with 72 new nodes with six core Intel Xeon processors using the Ivy Bridge microarchitecture. These new nodes have provided a substantial performance boost to LoBoS. In addition, 48 additional new nodes have been ordered, each of which will contain a Nvidia Tesla K20x GPU computing unit. Testing performed by the lab using CHARMM and other software have shown that GPUs are now viable for the labs molecular simulations workload, therefore the decision was made to invest in this new technology. Development of the CHARMMing web front end has continued. In the last year, focus was on enhancing the integrated lessons functionality of the interface. In addition, we have made enhancements to the on-line CHARMM tutorial (www.charmmtutorial.org). These led to the publication of a triple paper in PLoS Computational Biology. Continuing work with the H. Lee Woodcock group at the University of South Florida focuses on necessary infrastructure improvements and enhanced compatibility with non-CHARMM molecular simulation engines. Molecular simulation and modeling software packages are the vehicle for computational research and experiment. Implementation of new methods and options is the key to facilitate cutting edge researches. In recent years, this lab has developed a series new compuatational methods, such as the self-guided Langevin dynamics for efficient conformational searching and sampling, the isotropic periodic sum method for accurate and efficient calculation of long-range interactions, and the map-based modeling tool, EMAP, for electron microscropy studies. Implementation of these new methods enables researchers to tackle difficult problems. These methods have been implemented into CHARMM to expand its capability in molecular simulation, conformational search, and structure prediction. These methods are all available in CHARMM version 40. These methods are also been implemented into another widely used simulation package, AMBER, to extend the user scope to access these methods. The SGLD, IPS, and EMAP methods are available in AMBER version 14. Ongoing work has been focused on the development of replica exchange methods. Many of these efforts have centered around pH replica exchange, both with and without the use of reservoirs. The combination of Enveloping Distribution Sampling with Hamiltonian replica exchange (EDS-HREX) has provided a facile method for sampling different titration states of molecular system. One of the key benefits of this method is that it can be used effectively with explicit solvent, which removes a key source of error in many constant pH simulations. Also, a physically realistic conformation is sampled at each time step with EDS-HREX, in contrast with lambda-dynamics based methods. Current efforts involve expanding EDS-HREX to use reservoirs and two use two-dimensional replica exchange. Using two-dimensional EDS-HREX allows for conformational transitions between multiple pH values or temperatures, which dramatically increase the number of state changes. Initial testing has shown that this yields substantially faster state transitions than are achievable with EDS-HREX alone. Additional replica exchange based approaches focus in two directions. Firstly, A new method, Perturbed Reservoir Replica Exchange (P-RREX), is under development. The primary advantage of this method is that it does not make any assumption about the composition of the reservoir. Although sampling efficiency will be limited by reservoir quality, P-RREX will produce a Boltzmann ensemble using any reservoir. Continued testing of the early implementation has been performed. Secondly, work has been initiated on applying Suwa-Todo based global balance methods to replica exchange simulations in CHARMM. Pre-alpha level code has been developed and testing is starting. Structure Activity Relationship (SAR) and Quantitative SAR (QSAR) have been traditionally used in lead optimization approaches in drug discovery research to improve the probability of certain features such as activity, druglikeness, or absorption, distribution, metabolism, and excretion (ADME) properties of lead compounds. (Q)SAR procedures reduce costs of the early drug discovery pipeline,, and have a long history both in industrial design and regulatory assessment of pharmaceuticals., We have developed a Free Web tool for SAR and QSAR modeling based on open source technology to add to the services provided by CHARMMing (www.charmming.org). This new module implements 15 different advanced machine learning algorithms for example Random Forest, Support Vector Machine (SVM), Stochastic Gradient Descent, Gradient Tree Boosting etc. A detailed instructions for creating new models are available from www.charmmtutorial.org. A user can import training data to create new models (either categorical or numerical) in two ways: a) upload his or her own SD files which contain structures and activity information or b) use the Pubchem Bioassay search interface to query Pubchem database for relevant biological assay data via PUG REST and SOAP interfaces. , ,,,, For the first time we have provided users with a streamlined procedure for automatically downloading training data sets, as well as tracking the model generation process and running models on new data to predict activity. CHARMMing (Q)SAR tool automatically verifies new models by using well-known machine learning techniques such as cross-validation and y-randomization so users can immediately see whether the created model is able to calculate valid predictions. A user is presented with Area Under Curve (AUC) measurements for the training set, for the y‑randomized set, and an average AUC for 5-fold cross-validation for categorical modeling. Pearson correlation coefficient R2 is used for the same purpose for regression models. While there are other recently available tools such as OCHEM and Chembench, which offer web-based SAR model building, CHARMMing (Q)SAR service supports more extensive set of machine learning methods. It offers a user friendly interface developed to present the workflow in a straightforward way for novice and advanced users both and have the potential to dramatically change research strategies of academic research groups and non-profit organizations. CHARMMing (Q)SAR service can be used as a stand-alone utility or as a supporting filter for the docking procedure, or for challenges such as Tox21 data challenge and Teach-Discover-Treat initiatives.
Showing the most recent 10 out of 15 publications
