Simbios DBP and Seed Project recruitment: Process and Content

Simbios executive committee

Russ Altman, PI (

Scott Delp, Co-PI (

Jeanette Schmidt, Exec. Director (

Peter Lyster, NIH Program Officer (

Jennie Larkin, NIH Lead Scientific Officer (


Process and Schedule

Simbios was recruiting new Driving Biological Problems (DBPs) to start in the fall of 2007. Applications were due October 30th 2007 . We were specifically looking for Driving Biological Problems that are consistent and complement our core software development activities. We are interested in expanding our core software components and developing new application areas for existing software components, such as applications that expand the use of our multi-body dynamics component, Simbody. Descriptions of current DBPs and examples of other funded seed projects are given in the appendices. We were looking for DBPs based on NIH supported research, as well as seed projects that may eventually evolve to DBPs.


Preference was to be given to projects that:

  • address mesoscale modeling –hence filling the scale gap between molecular modeling and the modeling of organs, (see Appendix 1 for current DBPs).
  • complement core software development activities and are consistent with the physics-based modeling and simulation goals of Simbios.
  • integrate in a sensible way with systems biology and physics-based modeling
  • address multiscale modeling
  • have current NIH support for the underlying biology (such as R01, P-grant, etc.)


For DBPs: up to 10% support for the PI and 1 postdoc or software developer in support of the project for 1 to 3 years.

For seed projects: 1 a student or postdoc for six months



A 1-5 page description of the proposed work containing a:

  • brief background of the research, including current NIH support,
  • description of the work to be performed (specific aims),
  • description of the software contributions to be made available on (growth in the use of is a critical goal of Simbios),
  • summary of all personnel to be supported (or partially supported)

The Simbios Executive Committee and the Simbios Lead Science Officer (Jennie Larkin) and Program Officer (Peter Lyster) reviewed submissions.



1.      Posted on the Simbios website

2.      Direct mail to all Simbios collaborators – including

·        Simbios investigators

·        Active collaborators (R01, previous seed projects).

·        Investigators that have approached Simbios with regards to a collaboration

3.      Simbios Science Officers

4.      Simbios Science Advisory Board Members


Examples of currently and previously funded DBPs in order of scale

RNA Dynamics

PIs: Dan Herschlag and Russ Altman; NIH grant: P01 GM66275

Abbreviated version of Specific Aims:

Using the Tetrahymena ribozyme as a model, determine:

Aim 1: The mechanism of its rapid electrostatic collapse at the onset of folding

Aim 2: The spectrum of intermediates that are formed in folding and unfolding pathways

Aim 3: The “landscape” for its folding, preferred pathways, features critical for the choice of the pathways taken along this landscape under conditions that approximate physiological.


Myosin Dynamics

PI: James Spudich; NIH Grant: 5R01GM033289-20

Abbreviated Specific Aims:

Use experiments and simulations to elucidate the mechanism of energy transduction by the following myosin motors:

Aim 1: Myosin V: A classical myosin, with strong evidence for a lever arm mechanism.

Aim 2: Myosin VI: A non-classical myosin, which moves by an unknown mechanism.


Neuromuscular Dynamics

PI: Scott Delp; NIH Grant: R01 HD33929 and R01 HD 46814

Abbreviated Specific Aims:

Aim 1-3: Determine if, and under what conditions, the following contribute to excessive knee flexion: 1) Tight Hamstrings 2) Weak Hip Extensors or Weak Ankle Plantar Flexor

3) Tight Iliopsoas Muscles

Aim 4: Evaluate the Capability of Dynamic Simulations to Identify Appropriate Treatments.


Cardiovascular Dynamics

PIs: Christopher Zarins and Charles Taylor; NIH grant: 1R01 HL64327

Abbreviated Specific Aims

Aim 1: Utilize 4D MRI analysis methods to characterize hemodynamic conditions and wall motion of the porcine and human thoracic and abdominal aorta.

Aim 2: Apply the cardiovascular dynamics methods to construct a virtual aorta.

Aim 3: Modify the virtual aorta model to match the anatomic dimensions and wall thickness of the aortas of the normal volunteers imaged in the companion R01 grant.

Examples of previously funded seed projects

PI: Darryl G. Thelen, Ph.D., University of Wisconsin-Madison 

The Incorporation of Ground Contact Models When Simulating Subject-Specific Gait



Rehabilitation of persons with gait disorders is a significant and challenging health concern. The challenge arises, in part, because gait pathologies can have many sources, including neurological injury, muscle weakness, and contractures. Current motion analysis techniques, based on inverse dynamics, are often insufficient to identify the cause of an individual’s pathology (7). Forward dynamic simulation approaches overcome this limitation by directly representing the causal relationship between muscle excitations and movement (7). Furthermore, simulations can be used to predict the consequences of interventions such as strengthening, functional electrical stimulation and surgical treatment (1). In this work, we seek to enhance the capability and availability of software tools for simulating subject-specific gait, and for predicting the effects of interventions on movement. 


We have previously established a computationally efficient approach, termed computed muscle control ( CMC ), for generating muscle-actuated simulations of subject-specific motion (4, 5). With CMC , complex simulations are generated in minutes, which is two to three orders of magnitude faster than conventional approaches. More importantly, the simulations are extremely accurate with joint angles typically tracked within one degree. We have successfully used CMC  to analyze the causes of gait disorders (2) and to characterize muscle mechanics during running (6). However, a remaining shortcoming of our current approach is the direct use of measured ground reaction force trajectories in the forward simulation. While sufficient for studying small deviations in movement, this approach is insufficient to simulate the deviations in movement, which can occur after major interventions, such as orthopedic surgery. 


Description of Work to be performed 

The goals of this work are to: 

1) Develop an algorithm for identifying the parameters of a ground contact model that replicate subject-specific ground reaction forces. 

2) Incorporate the contact identification algorithm into a suite of software tools (uwpipeline) to be maintained on for efficiently generating simulations of human movement. 

Details omitted


Software Contributions to be made to SimTK 

Our algorithms will be coded in C and made compatible to link with neuromusculoskeletal models created using SIMM-Pipeline, which is the leading software package for simulating human movement. Source code, executable programs, examples and documentation will all be deposited in an open-source project (uwpipeline) on, so as to enhance the capability of others to adopt and use the framework. These software packages will also be compatible with OpenSim, since that framework is designed to be backward compatible with SIMM-pipeline. The eventual re-coding of the uwpipeline software as a plug-in for OpenSim will likely be warranted as the software matures and becomes useful to the community.

PI: Patrice Koehl, Computer Science and Genome Center , University of California , Davis

ElecTK: A Toolbox for Macromolecular electrostatics


Specific Aims:

Develop and implement a new formalism for computing macromolecular electrostatics. 


We propose to use a generalized Poisson-Boltzmann-Langevin equation to describe the electrostatic field generated by a macromolecule in water, based on a description of the solvent and ions as an assembly of freely orienting dipoles. We will develop an extensive computational toolbox for electrostatics calculations, ElecTK that will be incorporated into SimTK



Traditionally, the non-linear Poisson-Boltzmann equation (NLPBE) provides an accurate description of the electrostatics field around a molecule (for a recent review, see Koehl, 2006). The NLPB formalism assumes continuum solvent, with a fixed dielectric constant, as well as a bulk description of the ionic interactions (using Debye-Huckel theory). While NLPBE is used routinely to study electrostatic interactions in proteins, its applications to nucleic acids are more limited, as these two assumptions may then not be valid. Water is known to be organized in at least a first layer around macromolecules (the “water sites” in the case of tRNA (Westhof, 1988), and some ions are known to be immobilized at specific sites. There is therefore a need for a new formalism for the computation of electrostatics energy that includes the density of water dipoles as well as the density of ions around a molecule as parameters. Electrostatics calculations remain computationally intensive and difficult to set up, despite efforts to develop web interfaces to facilitate the latter (see Koehl, 2006). There is therefore a need for a comprehensive toolkit for computing macromolecular electrostatics that would be freely available, easy to interface with other applications, and available as a web service to allow non specialists to perform these types of calculation. The development of such a toolkit is the central aim of this proposal. 


Research Design



ElecTK: a toolkit for electrostatics calculations. We will develop a toolkit for computing macromolecular electrostatics that will include the fast PBE solver and the GPBLE solver described above. This toolkit will include all source codes required to implement these calculations in other programs, as well as web services to provide access to these resources to non specialists. ElecTK will be freely available under SimTK, to ensure broadest possible use by the biomedical community.

PIs: Robin Gutell and Pengyu Ren , UT Austin

Physical-based Coarse-Grain Modeling tool for RNA 



We propose to develop a preliminary version of physical-based coarse-grain RNA modeling program that will compliment Simbios’ effort in this area. 


Background in Molecular Mechanical Model for RNA Molecular mechanical simulations are used routinely to study bimolecular structure and dynamics. At the core of molecular mechanics is the empirical physical potential (force field) that provides the quantitative description of molecular interactions. The current major all-atom force fields for nucleic acids include AMBER and CHARMM. More and more successes in nucleic acids modeling have been reported as the force fields are constantly being improved and optimized [Cheatham, 2004; Mackerell, 204].

However, the current state-of-the-art of all-atom simulations is limited to nanoseconds and a few hundred nucleotides, whereas much of RNA dynamics occur on longer temporal and spacial scales. To overcome the time and size scale barrier, coarse-grain (CG) models where atoms are lumped into one particle have been actively sought after for modeling macromolecules. Bead-on-spring models have been attempted recently on DNA double helix [Tepper, 2005]. Mergell et al. proposed a model for DNA where each base-pair is represented by a Gay-Berne ellipsoid and the sugar phosphate backbone is modeled by harmonic springs. Harvey and coworkers have been utilizing coarse-grained molecular mechanics model with pseudo nucleotides to construct low resolution RNA structure. Simpler models based on experimental stacking free energy [Freier, 1986; Turner, 1987] and Gaussian chain entropy have been applied to predict RNA pseudoknots [Isambert, 2000; Xayaphoummine, 2003].


In general there are two types of approaches available for coarse-graining, referred to as knowledge- and physical-based. The former begins with direct survey of existing molecular structures. The later follows the physical interpretation of molecular interactions such as those represented in the all-atom force fields. In principle, a physically consistent model is more transferable; e.g. a RNA model can be applied to RNA-protein systems without significant reparameterization. Furthermore, models at various coarse-grain levels can be combined in a multiscale fashion to improve the computational efficiency while retaining critical details. Previously, we have developed point multipole based electrostatic framework for biopolymers (polarizable protein force field AMOEBA). It is our goal to extend this framework together with anisotropic vdW function to build up a coarse-grained molecular mechanics model for RNA.


Research Plan 



Software contribution to Simbios 

With the seed grant support, we will provide an initial version of physical-based coarse-grain RNA modeling program that will compliment Simbios’ current effort in this area. The program will consist of energy and force evaluation routines based on the coarse-grained physical model and associated parameters. This routine can be combined with SimTK molecular dynamics engine NAST. We will also provide a primitive rigid-body MD engine. With our internal structure manipulation program, visualization will be made through SimTK’s ToRNADo program. 

PI: Jill Higginson, University of Delaware

Simulation of post-stroke hemiparetic gait: 2d vs 3d 



Stroke affects approximately 700,000 Americans each year and is the leading cause of long-term adult disability. Post-stroke hemiparetic gait is complicated by impaired muscle activation, which results in abnormal muscle coordination patterns and asymmetric movements. An improved understanding of muscle function and dysfunction will facilitate the design of appropriate therapeutic interventions based on each patient’s needs. Through the use of muscle-actuated forward dynamic simulation, we can assess the relationship between muscle function and observed movement patterns to identify limitations and areas for improvement. Thus, the objectives of this short-term proposal are to generate subject-specific simulations of a range of gait patterns from individuals with post-stroke hemiparesis, compare the performance of two optimization algorithms, and identify any differences between two- and three dimensional simulation results. 


Specific Aims 

1. Generate subject-specific simulations of up to 10 subjects who walk with a range of post-stroke hemiparetic gait abnormalities (e.g. foot drop, equinus, hip circumduction). We will study how muscle excitation patterns are linked to observed function and how these differ between individuals. 

2. Compare performance of two optimization algorithms ( SPAN and CMC ) for 2d and 3d simulations with a representative case study. 

3. Identify differences in muscle function deduced from 2D and 3D simulations for each walking pattern. 



According to the American Heart Association, 7.7 million people are living with the effects of stroke and over 700,000 people will experience a stroke or recurrence of a stroke annually. Because stroke impairs walking function, it is a leading source of long-term adult disability. Gait deviations following a stroke are multiple and varied. Spatio-temporal abnormalities include asymmetric stance and swing duration, step length, push-off forces and decreased walking speed. Zajac asserts that the basic principles of movement coordination remain unclear despite years of collecting kinesiological data. Forward dynamic simulation, in conjunction with experimental measurements and optimal control theory, demonstrates promise at elucidating principles of muscle coordination. Muscle-actuated forward dynamic simulation offers a deterministic framework, which permits the identification of underlying muscle impairments and investigation of resultant movement abnormalities. Moreover, the potential effect of individual muscles on movement patterns can be studied, and used to deduce beneficial compensatory strategies and therapeutic interventions. 


Simulation studies have been used previously to identify unique and synergistic properties of the uniarticular and biarticular plantarflexors and other muscles during walking in healthy adults. Through investigation of a muscle-actuated forward dynamic simulation of normal walking, Neptune et al. (2001) found that soleus and gastrocnemius provide vertical support of the center of mass (COM) during stance phase, soleus contributes to forward acceleration of the trunk in pre-swing, and gastrocnemius generates pre-swing leg energy. Based on the same framework, Higginson (2005) developed a simulation of slow gait (0.3 m/s) in healthy adults to study the altered contributions of the plantarflexors to COM support. From this muscle-actuated forward dynamic simulation of healthy slow gait, we concluded that healthy individuals maintain COM support at slow speeds via minor modifications to the muscle coordination pattern associated with normal speed walking (i.e., increased contributions of knee extensors and enhanced stiffness due to ankle dorsiflexors). 


Higginson et al. (2006) subsequently developed the first muscle-actuated forward dynamic simulation based on the kinematic and kinetic data from a complete cycle of post-stroke hemiparetic gait. The simulation framework used previously was enhanced to permit asymmetric kinematics and muscle control. During gait for an individual with post-stroke hemiparesis, adequate COM support is provided via reorganized muscle coordination patterns of the paretic and non-paretic lower limbs relative to healthy slow gait (i.e., co-activation of paretic muscles). Higginson et al. (2006) have also used perturbation studies to investigate the effect of equinus foot placement, a common movement abnormality, on musculoskeletal dynamics. In a forward dynamic simulation of normal walking, we augmented ankle plantarflexion by 10° at initial contact and found that equinus posture alone (without concomitant changes in muscle forces) caused the knee to hyperextend, while intrinsic force-length-velocity properties of muscle diminished the effect of equinus posture alone. Our preliminary work is exemplary of our proficiency with experimental and computational techniques related to the study of muscle coordination in individuals with post-stroke hemiparesis


Research Plan



Software Contributions 

  • Simulated parallel annealing by neighborhood ( SPAN ) will be posted on This parallel implementation of the global search algorithm expedites time to convergence for large-scale simulations of human movement by at least a factor of 20 (Higginson et al., 2005). 
  • New simulations of post-stroke hemiparetic gait, which represent a range of gait abnormalities, will be posted with initial conditions, muscle excitation patterns and movies on