Molecular Dynamics for Everyone: A Technical Introduction to the Molecular Workbench Software

Transcription

1 Molecular Dynaics for Everyone: A Technical Introduction to the Molecular Workbench Software Charles Xie The Advanced Educational Modeling Laboratory The Concord Consortiu qxie@concord.org The world oves because olecules ove. Studying the otion of olecules is iportant to the understanding of any critically iportant concepts in physics, cheistry and biology. A fundaental goal of scientific research is to learn how things work, which at the icroscopic level priarily eans how atos and olecules ove to perfor certain functions such as cheical reactions, olecular recognition and protein synthesis. The ability to reason about coplex phenoena in ters of fundaental facts and theories describing the structures, interactions and dynaics of the atos and olecules of which all things are ade is called the Molecular Literacy [1]. This botto-up perspective, in addition to its philosophical significance, is increasingly iportant, as nanotechnology and biotechnology are eerging as the twin engines of the next industrial revolution. Molecular Literacy epowers people to better understand known phenoena and to anticipate the properties of novel arrangeents of atter under new conditions. Achieving Molecular Literacy at the grassroot level is therefore crucial for future scientists and engineers as it prootes the iaginative thinking needed to create and develop new nano-structured and bio-ietic aterials and devices for diverse applications to eet critical healthcare, energy, and environental challenges. Since the invention of olecular graphics, a subject that focuses on visualizing olecules using 3D coputer graphics, cheists have ebraced tools capable of rapidly displaying olecular structures and viewing the fro different perspectives. Because of the appealing effect of 3D graphics, coercial coputer-aided design packages often feature olecular visualization tools to proote the products. Since the advent of the Internet Era, several free tools have been developed for showing olecules on the Web. These tools are widely used by educators to teach olecules []. Most olecular viewers, however, are ainly designed to show static structures. The user can rotate and translate the entire rigid structure or change the view angle dynaically to create a otion effect, but the atos do not ove relatively to each other. Viewing static structures helps students learn the structures, but it is often far ore iportant to learn the functions after all, we study olecules because we hope to ake use of what they can do. Although one can argue that in any cases there exists a strong structure-function relationship that can help people derive functions fro structures, it is ipractical to expect inexperienced students to be able to reason using the relationship that ay be evident only to experts. Too often have we seen an excited cheist trying in vain to explain to nonexperts what he or she sees in a 3D odel of a olecule that is beyond a cool picture of soe 3D structure. It would be ost helpful if a olecular echanis can speak for itself in a dynaic visualization to fill the gap. Soe olecular viewers can sequentially display a series of static fraes, which can be different states of the sae olecule observed experientally or coputed theoretically, to create an aniation of a conforational change. This is a step forward to help students learn olecular echaniss, but it only allows the to passively watch what was set to happen. For a tool

2 to do a better educational service, students should be peritted to ess up with the odels, try any what-if s, and see what happens. Funded by the National Science Foundation, a tea of the Concord Consortiu has been developing a free, open-source progra called the Molecular Workbench (MW) ( which brings a salient, dynaic olecular world to the coputer screen and allows students to interact and experient with it 1 based on real-tie olecular dynaics (MD) calculations and visualizations [3]. MD odeling provides a powerful eans to foster the Molecular Literacy because it copleents and enhances traditional instructional approaches, including foral atheatics. Cognition can be viewed as a process of aking and anipulating ental odels of iaginary objects and events [4]. A scientific odel that coprises basic units of coherently structured knowledge in a tested and integrated fraework, if presented appropriately with effective pedagogy, can be enorously useful in helping students develop correct ental odels fro which they can ake logical inferences such as explanations, predictions and designs. Molecular odeling, which is an iportant part of conteporary scientific research [5, 6], constitutes the theoretical foundation for creating objective conceptual odels [7] that can be used by students to explain and investigate any natural phenoena at the olecular scale and thereby develop concrete ental odels about the. It is iportant to ephasize the educational significance of coputational odels based on science as opposed to ovies coposed of fraes of iages and aniations based on siple tieline rules. In a coputational odel, critical behaviors eerge fro algoriths derived fro first principles, which give it the explanatory power that can be used to anifest existing knowledge and the predictive power that can be used to explore unknown doains. A coputational odel can accurately siulate a large nuber of different phenoena by varying paraeters, configurations, initial conditions and boundary conditions. A ovie or an aniation, in contrast, can only illustrate a handful of situations that are recorded or preprograed. If a picture is worth 1,000 words and a ovie is worth 1,000 pictures, a odel is worth 1,000 ovies, or 1000,000 pictures, or 1000,000,000 words. Because a odel provides a uch larger intellectual space and ore freedo for learners to explore, create and invent, effective, profound learning is ore likely to occur. This article presents the scientific ethods, the educational background and the technical ideas behind the software. It is assued that you have soe preliinary knowledge about coputational science, educational technology and software engineering. In the first section, we review the basic procedures of classical MD siulations. In the second section, we discuss the requireents of educational siulations and the technical work needed to be done to eet these requireents. In the third section, we introduce our ideas to evaluate student learning based on MW aterials. The results of our educational research are not covered in this technical article. Please see a paper by Pallant and Tinker [8] if you are interested in the. How to siulate the otion of atos and olecules? Everything oves because of forces, which result fro the interactions aong atos and olecules. We will begin with how theoretical cheists odel interatoic and interolecular interactions. 1 You can watch a ovie about MW at:

3 There are two levels of odeling for olecular interactions. One is based on quantu echanics calculations [9], which is beyond the scope of this article. The other uses epirical fors, which will be introduced in the following. Molecular echanics force fields Atos basically interact with each other through van der Waals forces and electrostatic forces. When they are covalently bonded to others, strong forces hold the together as stable cheical groups. A widely used atheatical odel for the potential energy of a olecular syste consists of six types of interactions: U V LJ V EL V BS V AB V PT V IT The first type, V LJ, is the Lennard-Jones potential that has an attractive part representing the van der Waals energy and a repulsive part representing the Pauli repulsion: V 1 1 ij ij ij Rij Rij LJ 4 i, j, i j where R ij is the distance between the i-th and j-th ato, ε ij is called the van der Waals dissociation energy, and σ ij is called the collision diaeter. The dissociation energy is equal to the aount of energy needed to pull a pair of atos in the strongest van der Waals binding state apart. The collision diaeter is approxiately the distance at which a pair of atos bounces off fro each other in a noral, non-reacting condensed state. The power of the negative ter, which is soeties also called the London dispersion force, has a root in the quantu echanical calculation of the binding energy of the hydrogen olecule, but the power of the positive ter has no apparent theoretical basis (soeties, it is set to be 9 to soften the repulsion core for dense systes). V EL is the electrostatic potential energy according to Coulob s Law: 6 V EL 1 i, j, i j q q i R ij j where q i is the charge of the i-th ato. Copared with the van der Waals potential, the electrostatic potential is a stronger, ore long-range interaction. A pair of charged atos in vacuu will be able to feel each other fro quite a distance away, whereas a pair of neutral atos will feel each other s existence only when they are close. For crystals or solutions, the Ewald Su is often needed to copute the suation of the weak contributions fro nuerous reote charges. But if only qualitative results are needed, this expensive procedure ay be skipped to speed up the siulation. V BS is the bond-stretching energy standing for the elastic interaction between a pair of atos connected by a covalent bond, V AB the angle-bending energy standing for the interaction aong three If a pair of atos can react to for a covalent bond, the length of the bond between the can be saller than the collision diaeter.

4 covalently-bonded atos that for a stable angle, and V PT and V IT the proper and iproper torsional energies standing for the interactions aong four covalently-bonded atos that for a stable proper and iproper dihedral angle (see Fig. 1): V V V V BS AB PT IT bonds angles torsions k k k l torsions ( l l 0 ) 0 ( ) 0 ( ) V [1 cos( n )] where l is the distance between Figure 1: A scheatic illustration of the interactions that odel covalent bonding: (A) Bond-stretching force; (B) Angle-bending force; the two atos of the -th bond, l 0 is the equilibriu bond length, k l (C) Proper torsional force; (D) Iproper torsional force. is the bond strength, is the -th angle between the two adjacent bonds that share a coon ato, 0 is the equilibriu bond angle, k is the strength, is the -th dihedral angle between the two adjacent angles that share a coon bond, n is the periodicity factor which deterines the nuber of equilibriu dihedral angles in a 360º rotation, is the phase shift, V is the aplitude, ξ is the -th iproper dihedral angle aong four atos that are not bonded successively to one another, ξ 0 is the equilibriu iproper dihedral angle, and k ξ is the strength. 3 The last four ites are called the bonded interactions, which aintain the bond lengths, the bond angles and the dihedral angles so that cheical groups will reain sterically stable in an MD siulation 4 (in MW we call these constructs radial bonds, angular bonds and torsional bonds). The first two ters, the Lennard-Jones potential and the electrostatic potential, are called the non-bonded interactions. In MD siulations, they are ore iportant than the bonded interactions. It is the nonbonded interactions aong the atos of a acroolecule that affect its secondary structure. It is the non-bonded interactions aong the atos of different olecules that organize the into crystals, coplexes and other asseblies. It should be pointed out that the decoposition of the potential energy of a olecular syste into the above force field ters is purely epirical. In other words, they are the atheatical odels for describing the cheical forces that stabilize the structures derived fro diffraction patterns obtained 3 The bond-stretching potential given by Hooke s Law does not perit a bond to break. The ore a bond is stretched, the greater is the force to pull the atos back. As a result, the above force fields cannot be used to siulate cheical reactions, which involve aking and breaking bonds. We have proposed a ethod that allows bonds to ake and break, and thus akes it possible to siulate soe siple reactions [10]. 4 Soe olecules, such as benzene, have delocalized bonds that involve ore than four atos. However, no higher ters of energy decoposition see to be necessary in this treatent.

5 using crystallography. The force field approach per se has no power in explaining how the structures are initially assebled fro discrete atos, which is essentially a question of how to siulate the origin of life. Most MD packages for bioolecular siulations are based on the olecular echanics as described above, though they ay differ in the paraeterization (the deterination of the bond strengths, the van der Waals paraeters and the partial charges). In MW, to ensure broad applicability, an ato can be considered as a generic particle with paraeters that can be changed freely. Molecular dynaics siulations Having defined the interactions aong atos, the position, velocity and acceleration of each ato are calculated using a nuerical ethod (e.g. the Verlet ethod or the Runge-Kutta ethod) to solve Newton s equations of otion according to the forces derived fro the gradients of the interaction potentials involving the ato: ir i iu R1, R,..., R where R i is the position vector of the i-th ato and i is its ass. The nueric integration is carried out stepwisely. The process is repeated at each discrete tie step. The trajectory of each individual object can be tracked by connecting its states to for a tie series. The tie evolution of the entire syste can be viewed as a fiber bundle of tie series in the phase space. That is all you need to do to get an MD siulation up and running. For advanced topics, such as boundary conditions, therostats, pistons, statistical analysis and so on, interested readers can consult with Ref. 6, or read the online User s Manual within MW. The MD ethod is very useful in scientific research, because it satisfies the following fundaental physical laws: n The First Law of Therodynaics. The Law of Conservation of Energy autoatically eerges in an MD siulation. If there is no energy input/output through external forces or dissipation through friction, the total energy, which is the suation of the potential energy and the kinetic energy for all the atos in the syste, reains constant within the tolerance of nuerical errors. This can be used as a criterion to check if a siulation runs properly. The Second Law of Therodynaics. Although the Reversibility Paradox suggests that classical dynaics ight be at odds with the Second Law of Therodynaics, MD siulations of basic processes such as diffusion, heat transfer and phase transition clearly show that the entropy of an isolated olecular syste always tends to axiize. Despite of the fact that it is possible to create special initial conditions that lead to a process of entropy reduction in an isolated syste, 5 in practice we have never found that such special conditions can spontaneously arise during an MD siulation for a any-body syste. 5 Consider an ipact process in an isolated syste: a high-speed ato bobards and breaks a icrocrystal. If we stop the siulation after the crystal has been broken, reverse the velocities of every single ato in the syste, and then continue to run it, we can reverse the process the atos re-asseble into the original crystal. The entropy decreases in this spontaneous process in

6 The Law of Moentu Conservation. As the Law of Conservation of Energy, this Law also autoatically eerges in an MD siulation. This Law dictates each collision aong atos. The overall result is that the total linear and angular oentu of the syste conserves. Other statistical laws. Iportant laws in statistical echanics, such as the Theore of Energy Equipartition, Maxwell s Theore of Speed Distribution, and the Boltzann Distribution are all guaranteed in MD siulations. We can even siulate the Galton Board that deonstrates the noral distribution. 6 What needs to be done to ake olecular dynaics odeling accessible to students? Most MD progras involve using a pre-processor to prepare siulations and a post-processor to analyze results. When a calculation is actually being done after it is subitted to a coputing service, the user is rarely given a chance to intervene. Moreover, any progras require the user be able to work with coand lines and scripts, and feel cofortable dealing with raw data. These requireents are prerequisites for scientists. But they becoe disadvantages when novice users in schools try to use the without the aid of an expert. The overarching goal of a learner-friendly MD progra is that average students can use it to learn science. The interingled coplexity of learning and science requires a highly integrated syste that is capable of supporting both. Instead of having students work with a set of distinct tools and switch back and forth, the ideal technological solution encapsulates of the entire process of building odels, setting up conditions, running and controlling siulations, visualizing results, recording observations, testing, onitoring learning progress, and feedback into a single progra with a unified graphical user interface. In the following subsections, we will discuss the iportant facets of such a syste in details. Siulations ust be interactive When a scientist perfors an MD siulation, the goal is not to watch how it unfolds on a coputer screen. There is seldo a need to spend precious coputing resources on visualizing the interediate results on the fly. Neither is there a need for the user to alter the paraeters and conditions arbitrarily during a siulation. Often, a siulation starts with a fixed set of inputs, and records the interediate results while it is running. When it copletes, the stored results can be analyzed to retrieve the needed inforation corresponding to the given set of inputs. For a progra to be ore educationally useful, however, opportunities ust be provided to students to interact with siulations theselves. To support inquiry, students ust be allowed to adjust paraeters and add inputs at any tie, and see the eergent behaviors of the siulated systes the siulation without work and/or cooling fro the outside world. This paradoxical result sees to be a violation of the Second Law of Therodynaics at first glance. But in practice, the probability of encountering the state of a any-body syste in which all velocities take the values and directions as if they were anually reversed as described above is so low that it can be neglected. See: 6

7 instantaneously. Only through interacting with siulations freely in any different ways and watching the results can students discover the cause-and-effect relationships revealed by the siulations and therefore construct their own ental pictures about the iportant physical and cheical concepts ebodied in the siulations. Technically, an educational siulation is required to do both the calculations and the visualizations at the sae tie in order for students to see the entire process in all possible levels of details, and in real tie in order for the to anipulate the syste and observe its responses right away. Translated into prograing ters, the MD code, the visualization tool and the graphical user interface ust be integrated sealessly in the run tie. The requireent of interactivity, however, liits the sizes of siulatable systes on personal coputers. There are two solutions to this issue. The first one counts on the continuous iproveents on coputer power. As the Era of Multicore Coputing is upon us, parallelizing the software syste to axially use the power of ulticore processors will allow larger systes to be siulated. The second one uses a coarse-grained approach to reduce large systes into odels with a tractable nuber of particles, each of which represents a large nuber of grouped atos that for a stable structure with insignificant internal vibrations. These effective particles interact with each other to ove, join and break. For exaple, it is coon to use a odel in which an aino acid is represented by a single particle to study the echaniss of protein folding [11]. Iplicit solvation that eploys effective fields to siulate solvent-solute interactions can be used to save the coputational cost needed for the vast nuber of water olecules that have to be otherwise present in the siulations of olecules in aqueous solutions. The blue window in Fig. shows a coarse-grained odel for icelle with iplicit solvation. Siulations should be easy to create Models are abstractions of data that are coposed of different objects at different locations in the phase space under different conditions and with different initial settings. A siulation engine, if generic and powerful enough, allows us to build as any odels as peritted by its capacity to atch the diversity of reality. It becoes apparent that, to harness the power of this theoretical capability, users need to be able to turn their odeling ideas into coputational odels. A userfriendly syste for constructing siulations has twofold iportance. First, it allows educators to create siulations for teaching, as alternatives to traditional drawings Figure : A screenshot of the D Model Builder in action. The enu bar and the tool bar above the view window provide any tools needed to construct odels and set up siulations. Each type of object also provides a pop-up enu and a property editor fro which the user can edit and odify the properties of an object of its type.

8 and illustrations. Second, it allows advanced students to design their own odels, a process during which their odeling skills and creative thinking can be trained. Engaging students to construct odels and siulations ay also serve as an introduction to olecular odeling, which will benefit the if they end up choosing science and engineering as their future careers. MW has a What-You-See-Is-What-You-Get (WYSIWYG) D odel builder that akes odels in a way that is as easy as aking shapes with a drawing progra (see Fig. ). With this odel builder, any types of objects can be added, and every object in a odel can be cut, copied, pasted, draggedand-dropped, and edited through the supporting pop-up enus and property editors. User actions are undoable and redoable. Annotations can be added to ake a odel easier to understand. For a coarse-grained particle odel, custo iages can be attached to decorate the particles so that the appearance of a odel will bear a reseblance to illustrations coonly seen in textbooks, particularly for olecular biology (see the siplified graphical representation of lipid olecules that for the icelle in Fig. ). Thanks to the integration of siulation and construction, there is no border between construction and run in MW. The user can, at any tie, run the odel while constructing it. This characteristic feature, steing fro the dynaic nature of MD odels, is a ajor difference between building a static odel and building a dynaic one. In fact, test-running a odel under construction is an iportant part of the constructing process as it allows the user to build through trail-and-error cycles. Unstable constructs can be autoatically reoved or spotted when a odel runs (or a procedure of energy iniization is called). For users who are not satisfied with the abstraction of D odels, a rudientary WYSIWYG 3D odel builder is available for creating 3D odels (Fig. 3). It allows the user to build olecules fro scratch by laying down atos in the 3D space with the assistance of ovable helper planes and joining the by radial, angular and torsional bonds. More coplex cheical systes can be built using a set of building blocks that includes all the aino acids and nucleotides and any sall organic olecules. A crystal builder is also provided to build a liited nuber of crystal lattices. Atos can be Figure 3: A screenshot of the 3D Model Builder in action. It can be used to build odels as coplex as this nano car, which has four short carbon nano tubes as the tires and wheels. selected, translated, rotated, duplicated, and deleted as blocks. Three different views are available for the author to set the perspective to observe a siulation. These views include a regular view in which the odel is viewed as a whole at different zooing distance, a navigation view in which the user

9 can ove the caera around to experience an iersive effect of flying into a olecule, and a rover view in which the caera is attached to an ato to iic the effect of riding on an ato when a siulation runs. There are any pre-ade odels that cover quite a breadth of science and are freely available in MW. The collection constitutes a solid scientific foundation of MW. As the software developent continues, ore exaples will be added to consolidate and expand this foundation. Siulations should be ebedded in a learning environent Siulations are not broadly useful in classroo without accopanying instructional aterials. Many educational applications provide lesson plans or worksheets for students to use, separated fro the software tools. But the optial way of using siulations is to ebed the in a learning environent that provides all the essential eleents needed for a learning process. We call such a coplete package a learning activity. A good learning activity otivates, scaffolds, and supports student exploration of odels and siulations. It also provides background inforation, opportunities for reflection, and ethods of onitoring student progress in a designed context and evaluating learning. MW is a versatile learning environent that offers this kind of classroo-ready learning activities. Moreover, it provides an authoring syste for creating the. A learning activity in MW usually consists of ultiple pages. A page is a screen space in which text can be typed and styled like in a word processor and any kinds of coponents can be inserted and custoized. The fact that these eleents can be placed anywhere on a page and ixed with characters, iages and links allows the author to design high quality, visually appealing and selfexplanatory pages. Authors who have experience in creating siple HTML web pages should be able to author siilar pages in MW without a proble. Figure 4: This screenshot shows that an MD odel for the Brownian otion is custoized and contextualized in an activity that teaches the concept of scale. A odel container, within which a siulation runs, is a core coponent that can be inserted into a page. With a rich set of pluggable coponents that can counicate with a odel container through coand and data channels, a custo user interface can be built for each siulation. The user interface can coprise controls of the siulation, buttons and sliders for changing the paraeters, and graphs for displaying the outputs. Custoized user interfaces are iportant because they establish a learning space that is constrained only within the topics covered in the activity. For exaple, Fig. 4 shows a custoized user interface for showing the Brownian otion. Although standing behind the scene is the entire engine that is capable of doing nuerous other kinds of siulations, the end user of the activity needs not be concerned with anything beyond what is presented on this succinct page in order to learn the intended subject.

10 Another kind of coponents that can be inserted into a page is questions. There are three types of questions that the author can set up: ultiple choice question, free response question and iage question. An iage question is a type of question invented in MW, which requires the student to take a snapshot iage of a siulation as an answer to the question. These questions can be used to test the student s prior knowledge (pre-test) and easure the gains after learning through an activity (posttest), for exaple. The authoring syste, along with the odel builders, has allowed us to create a yriad of learning activities that can be found in the aterial repository within MW (the internal MW page when you press the Hoe Button on the tool bar of MW). Soe of these activities have becoe reasonably sophisticated and self-contained enough in content to be qualified as interactive textbooks. For those who are interested in learning how to create odels and activities, a coprehensive online User s Manual, which is written using the sae authoring syste, provides nuerous working exaples. Siulations should be easy to share A pedagogy ade possible by the easy-to-use odel builders is to involve students in creating siulations, a process that can be devised to ebed instructional steps that lead to progressive conceptual understanding. This learning path, different fro inquiry using interactive edia based on existing siulations created by experts, sees to be practical at college level where students have obtained adequate prior knowledge needed to understand the basics about siulations to get started. Fro the point of view of social constructivis, the creation process and the end product ust be shared with others in order for the full effect of learning to take root. It is through the creation of a olecular odel that is shared and becoes what Papert calls a "public entity" that learning is strongly reinforced [1]. In the process of building, sharing or collaborating, students learn their subjects well because they have to think hard about the and figure out how to present their odeling ideas to others. There is no hurdle for students to share a siulation in MW. The MW syste allows users to subit their odeling work through the MW Space, a web application that facilitates social interactions in virtual classroos in a culture where odels are the central eleents. Students can easily upload their odels to their MW accounts, and decide with who they will be shared. They can choose to share with their classates, their teachers, or the world users of MW. When the user opens an MW page that contains a siulation, it is ready to run. No extra step is needed. All saved paraeters, configurations and initial conditions are restored in a snap when it is loaded. With the authoring syste, students can ake procedures that should be followed to produce the desired results, and the controls to achieve the, as they would with real experients. They can also articulate their otivations, explain what their siulations will show, and narrate how they were constructed. This akes their work read ore like presentations instead of just plain odels that ay be less coprehensible. How do we know if students learn

11 To soe extent, going through an MW activity for a student is siilar to going through an experiental procedure in a wet lab. The coon things are that a student needs to read instructions, follow certain procedures, operate soe instruents, observe what happens, record soe data, and write a report at the end. The Report Syste in MW onitors what students do during this process, autoatically generates reports, and allows teachers to track down students progresses. Page 1 Page Page 3 The end Questions Questions Questions Report Generator Data Data Data The Data Collector and Storage Analysis Tools The MW Space Student Account Feedback Feedback The MW Space Teacher Account Figure 5: The flow of collecting student work during a learning activity, supported by the Report Syste. The Report Syste consists of five parts. The first one is questions, which can be ebedded into a learning activity as described in the previous section. The second is a data collector that gathers and stores a student s inputs fro questions. The third is a snapshot facility that allows students to take a snapshot iage of a siulation or a graph. A snapshot iage captures what a student sees happening on the screen, which are soeties difficult to describe erely with words or nubers. There is also a set of tools for annotating a snapshot iage, which students can use to highlight and explain certain parts of the iage. The fourth is a report generator that autoatically converts student data into a readable page that can be printed, saved or subitted. The fifth is the MW Space that registers students and teachers, receives and stores reports in a database, and provides feedback to teachers. Fig. 5 shows a scheatic illustration of the workflow of the Report Syste. The iportance of the feedback loop aong aterial developers, students and teachers ust be stressed. Producing learning aterials, designing effective pedagogies, and ipleenting the in classroo are essentially a dynaic, tie-consuing trial-and-error process that deeply rests on constructive interactions aong the users, the ipleenters and the developers. It is iportant that the teachers be able to track the learning processes of the students, and report probles and inefficacies to the developers. The aterials can therefore be revised and the quality be iproved.

12 With the Report Syste and its future iproveents on critical issues such as data ining and data analysis, MW can be a very useful tool for conducting educational research, particularly for studying the effectiveness of using coputational odels in education. While few disagree upon the extensive use of interactive edia in teaching, any (including us) are still searching for the best design strategy of these edia and the best pedagogy of using the. The research on this avenue ay yield insights that would develop the next generation of educational edia. Acknowledgeents: The author thanks B. Berenfeld, D. Daelin, D. Markan, A. Pallant, E. Rosenbau, B. Tinker and R. Tinker for their encourageent and nuerous suggestions for developing and iproving the Molecular Workbench Syste. This article is based upon work supported by funding fro the National Science Foundation. References: 1. Molecular Literacy, M. Patrick, unpublished anuscript.. World Index of Molecular Visualization Resources, 3. R. Tinker and Q. Xie, Applying Coputational Science to Education: The Molecular Workbench Paradig, Coputing in Science and Engineering, in press, D. Hestenes, Notes for a Modeling Theory of Science, Cognition and Instruction, paper presented at The 006 GIREP conference: Modeling in Physics and Physics Education, Asterda, T. Schlick, Molecular Modeling and Siulation, 1 st Edition, Springer, A. R. Leach, Molecular Modeling: Principles and Applications, nd Edition, Pearson Education, D. Hestenes, Modeling Theory for Math and Science Education, paper presented at the Matheatical Modeling ICTMA-13: Education and Design Sciences, A. Pallant and R. Tinker, Reasoning with Atoic-Scale Molecular Dynaic Models, Journal of Science Education and Technology, 13, (004). 9. D. Marx and J. Hutter, Ab initio Molecular Dynaics: Theory and Ipleentation, in Modern Methods and Algoriths of Quantu Cheistry, J. Grotendorst (Ed.), John von Neuann Institute for Coputing, Jülich, NIC Series, Vol. 1, Q. Xie and R. Tinker, Molecular Dynaics Siulations of Cheical Reactions for Use in Education, Journal of Cheical Education, 83, (006). 11. H. Jang, C. K. Hall, and Y. Zhou, Assebly and kinetic folding pathways of a tetraeric betasheet coplex: Molecular dynaics siulations of siplified off-lattice protein odels, Biophysical Journal, 86, (004). 1. S. Papert, Situating Constructionis, in I. Harel and S. Papert (Eds.), Constructionis, Ablex Publishing Corporation, Norwood, NJ, 1991.