Visualization of a Virtual Caenorhabditis elegans in WebGL

Andoni Mujika, Gorka Epelde, Alessandro De Mauro and David Oyarzun

Vicomtech-IK4, Mikeletegi Pasealekua,57, 20009, Donostia-San Sebastian, Spain

Keywords:

WebGL, Visualization, Physically-based Modeling, Caenorhabditis elegans.

Abstract:

This paper presents the work that has been done in Si elegans project in order to visualize the locomotion

and the behaviour of a virtual reproduction of the nematode Caenorhabditis elegans, one of the most studied

animal in neuroscience. The project aims to develop the ﬁrst hardware-based computing framework that will

accurately mimic this worm. It will enable complex and realistic behaviour to emerge through interaction

with a rich and dynamic simulation of a natural or laboratory environment. In order to visualize the physical

behaviours that emerge from the neuronal system that has been constructed in the project, a web environment

has been designed where the user will be able to deﬁne an assay and to run it in a WebGL-based 3D virtual

arena. For that a relation has been deﬁned from the physics based simulation (run on the server side) and the

simpliﬁed web rendering of it.

1 INTRODUCTION

The Si elegans project aims to emulate the neuronal

system of a nematode called Caenorhabditis elegans

(C. elegans) (Altun and Hall, 2009). The behaviour

that emerges from this emulation will be simulated

in a virtual arena, in three dimensions. C. elegans is

one of the best known organisms in the world and it

is widely used, since its genome is completely known

and its manipulation for chemical and genetic tests

is relatively easy. Moreover, the relative simplicity

of its neuronal system (the hermaphrodite C. elegans

has 302 neurons) and its rich behavioural repertoire

(e.g. locomotion, feeding and even certain social be-

haviours) make this nematode a perfect organism to

study and learn from, in the way to understand more

complex organisms.

Despite the fact that the connectivity of those 302

neurons is known, the knowledge of the scientiﬁc

community does not cover every behaviour generated

by the neurons and their effects in the worm. That is

why Si elegans project aims to provide the scientiﬁc

community with an emulation alternative to the labo-

ratory experiments for their research, accelerating and

enhancing the research process.

For that, a web-based three-dimensional virtual

arena is being constructed where the user will deﬁne

the environment (obstacles, air or liquid, etc.), and the

behavioural experiments that will be emulated (e.g.

touching the worm, placing food or toxic in the plate,

applying vibrations to the whole environment). The

experiment deﬁnition will be transferred to a set of

Field Programmable Gate Arrays (FPGAs) where the

neuronal emulation will be held. The neuronal net-

work will process and compute the behaviour of the

worm and a physics engine will calculate the locomo-

tion of the worm based on it. Finally the physics sim-

ulation results will be rendered in the web browser.

Additionally, the web-based user interface will dis-

play detailed results of the experiment at neuron, mus-

cle and the environment level.

This paper describes the work that has been done

in order to connect the web browser and the physics

engine so that the user will be able to deﬁne the assay

easily and will visualize it in a 3D environment. In

the following, section 2 reviews the state of art on the

C. elegans locomotion simulation and its visualiza-

tion. Section 3 and section 4 describe the physics en-

gine and the web-based interface and ﬁnally, section

5 draws conclusions and discusses the future work.

2 RELATED WORK

Since early nineties, several works that try to simu-

late the locomotion of C. elegans have been presented.

The simulations are becoming more and more realis-

tic, but there is a lot of work to do in order to achieve

a perfect reproduction of the worm.

Most of the works that have been done in this

164

Mujika A., Epelde G., De Mauro A. and Oyarzun D..

Visualization of a Virtual Caenorhabditis elegans in WebGL.

DOI: 10.5220/0005168301640168

In Proceedings of the 2nd International Congress on Neurotechnology, Electronics and Informatics (-2014), pages 164-168

ISBN:

 2014 SCITEPRESS (Science and Technology Publications, Lda.)

ﬁeld (Suzuki et al., 2005) (Bryden and Cohen, 2008)

(Voegtlin, 2011) follow the pioneering work by

Niebur and Erd

os (Niebur and Erd

os, 1991), regard-

ing the shape of the worm and the forces that are ap-

plied on it to obtain the locomotion. They construct

a 2D reproduction of the nematode with straight seg-

ments that represent the muscles. Different forces are

applied to those muscles to make the body move in a

similar way to the real worm: internal pressure of the

worm, elasticity of the cuticle (”skin” of the worm),

muscle forces and environmental forces.

Regarding 2D simulation of locomotion, Boyle et

al. (Boyle et al., 2012) have been developing one of

the most advanced models, where the behaviour of

the virtual worm is very similar to the real one in

different environments (water, gelatin or agar). In-

deed, they showed that the different locomotion of the

worm does not emerge from the worm, i.e. its muscles

work equally in any environment but the environment

makes C. elegans move in different ways. They also

constructed a neuronal model that makes the worm

adapt to the environment with a minimal sensory in-

put.

In recent years, there have been some efforts to

upgrade previous 2D works to 3D worlds. Their aim

is to offer a perfectly realistic (to go beyond a sim-

ple representation in two dimensions) reproduction

of C. elegans and a easier way to explore its body.

The model presented by Mailler et al. (Mailler et

al., 2010), using a limited neural model and a worm

made by 25 rigid cylindrical sections, obtain a sim-

ilar movement to those obtained by above described

models, but in three dimensions.

The main active project in the ﬁeld of simulation

of C. elegans is the OpenWorm Project (OpenWorm,

2014). In this case, the physical model is based on

the work of Palyanov et al. (Palyanov et al., 2012),

where the virtual worm is composed of point masses

and springs to model skin and muscles. The neuronal

system of this model uses the 23 neurons that han-

dle forward locomotion (Chalﬁe et al., 1985). At the

moment, in the OpenWorm project they are integrat-

ing Predictive-Corrective Incompressible Smoothed-

Particle Hydrodynamics (PCI-SPH) (Solenthaler and

Pajarola, 2009) to model the behaviour of all the parti-

cles that take part in the simulation (including ﬂuids).

The realism they want to achieve is very high, but the

performance is too slow for the moment.

In regard to visualization of the results obtained

by the physical engine described above, the Open-

Worm Project has made two different approaches.

First, they developed a WebGL-based browser (Open-

WormBrowser, 2014), where the user can explore a

3D reproduction of C. elegans with all its cells (mus-

cles, neurons and other organs). However, there is

no locomotion in this browser. On the other hand,

they constructed a web version of their physical en-

gine Geppetto (Geppetto, 2014), where the user can

launch some simple particle simulations and watch

some neuronal information.

Our work is similar to the OpenWorm project, but

there are some differences. Firstly, the neuronal sys-

tem of OpenWorm is software-based and in Si ele-

gans, it is implemented using FPGAs. Regarding

physics simulation and its visualization, we aim to

obtain a better balance between realism and perfor-

mance, i.e. we want to obtain a more realistic simu-

lation than other works in the state of the art, without

losing usability because of slowness. On the other

hand, we can say that we are closer to a complete

WebGL-based simulation of the locomotion of C. el-

egans that OpenWorm.

3 PHYSICS SIMULATION AND

WEBGL IMPLEMENTATION

In this section, we describe the physically-based

model that has been deﬁned to make the virtual worm

move forward. Besides, we describe the method

we follow to take this complex simulation to a web

browser.

3.1 Physics Engine

Our physically-based model (Mujika et al., 2014)

makes a simpliﬁcation in the muscle structure of C.

elegans. Usually, the worm has 95 muscles, divided

in 4 quadrants of 24 muscles, except for the ven-

tral left one which has 23 muscles. However, in

our approach we have considered 96 symmetrically-

distributed muscles, as most works described above

do.

Figure 1: Structure of the virtual C. elegans in the physics

engine.

In our case, biphasic springs are used to model

muscles. A biphasic spring is a regular linear spring

that changes its spring constant at a certain length.

VisualizationofaVirtualCaenorhabditiselegansinWebGL

165

This way, changing the stiffness at a certain point

we stop the elongation of the spring at certain point.

Moreover, in order to generate the locomotion, those

springs must be active, i.e. they must contract and

relax themselves without any external force. That is

why the signal that comes from the neurons (imple-

mented in the FPGAs) is inserted in the equation that

rules the behaviour of the springs.

On the other hand, the rings that compound the

body of the worm are represented with a square of

biphasic springs and are linked to other rings with

the muscles. This springs are stiffer and maintain the

body shape of the animal. Fig. 1 shows the structure

that has been constructed to simulate the worm. In

each quadrant, there are 24 red segments that repre-

sent the muscles and each ring has 4 green segments.

The ﬁrst and the last ring are smaller in order to obtain

a more similar shape to the shape of the worm.

Once the structure of the worm has been deﬁned

the forces that will be applied to the masses that are

located in the vertices of the structure must be deﬁned.

• As stated before, the elasticity of the cuticle of the

worm and the activation of the muscles is obtained

by the use of biphasic springs.

• Usual gravity is applied to all the masses.

• As gravity would collapse the whole structure, the

internal pressure of the worm must be simulated.

For that, a force ﬁeld that creates a similar effect

to a cube made of silicon is deﬁned. The body of

the worm is divided into a grid of hexahedra and a

Finite Element Method (FEM) is used to simulate

the opposition to collapsing tendency of the spring

structure.

• Friction is the key that makes the worm go for-

ward and in our model it is based on Signorini’s

law and Coulomb’s friction law. In each time step

the colliding points (including those of the worm

and those from the ﬂoor) are computed. Then,

based on the values of forces in the previous time

step, an iterative process is done to obtain the ac-

tual values of the contact forces.

• Collision with obstacles is detected and the resul-

tant forces are applied to the nematode.

All the physics simulation is run using the Simu-

lation Open Framework Architecture (SOFA), (Allard

et al., 2007). It is an Open Source framework primar-

ily targeted at real-time simulation. It offers a scene

graph editor for an easy creation of complex environ-

ments and it has a wide range of components (solvers,

objects, collision detection algorithms, etc.), includ-

ing the possibility to create new ones.

3.2 WebGL Implementation

The simulation of the locomotion of C. elegans is car-

ried out in a server using SOFA, but it must be trans-

ferred to the web browser. Since most users will be

interested in the behaviour of the worm (and not in the

muscles themselves), we use a realistic 3D reproduc-

tion of the nematode (see Fig. 2) that will be trans-

formed accordingly to the simulation in the physics

engine. To render the 3D environment without any

plug-in installation, Three.js library (Three.js, 2014),

that makes use of WebGL (WebGL, 2014), has been

used.

Figure 2: 3D reproduction of C. elegans.

Besides rendering the virtual nematode, its move-

ment must be synchronized with the locomotion of

the worm in the server. For that, animation bones have

been set along the body of the virtual worm. These

bones will be transformed accordingly to the mass

points of the physics model. The triangle mesh that

represents the body is attached to these bones, so the

mesh moves accordingly to the transformations of the

bones. Fig. 3 shows the positions of the 12 animation

bones in respect to the virtual C. elegans.

To synchronize the transformations of the 12 an-

imation bones with the muscle movements generated

in SOFA, our implementation of SOFA generates a

data stream with the positions of the 96 muscles. The

WebGL-based 3D environment takes this data and ap-

plies it to the animation bones. Speciﬁcally, the posi-

tion of each bone is set to the center of each second

ring of the worm in SOFA.

Figure 3: Simulation bones of the worm.

NEUROTECHNIX2014-InternationalCongressonNeurotechnology,ElectronicsandInformatics

166

Figure 4: 3D simulation of C. elegans in web.

4 VISUALIZATION

The web browser of Si elegans will have a part where

the user will deﬁne all the aspects related to the ex-

periment he wants to carry out. This part will have

three main windows: the environment deﬁnition win-

dow, the experiment deﬁnition window and the 3D

window.

In the environment deﬁnition window the user will

set all the parameters regarding the habitat of the

worm. Some parameters will be discrete (deﬁned with

a dropdown menu) and other continuous (deﬁned with

a numerical value or a slider):

• Shape of the plate (discrete).

• Environmental substance. Discrete (water, air, sil-

icon, etc.) or continuous (density value).

• The initial position of the animal (continuous).

• Friction of the plate. Discrete (different materials)

or continuous (friction value).

There may be some aspects that must be deﬁned

using both the environment deﬁnition window and the

3D window. For example, to insert obstacles in the

plate, the user will click in the environment deﬁnition

window to choose the obstacle type and then click in

the 3D window to set the place where the obstacle will

be.

On the other hand, the experiment deﬁnition win-

dow will be used to deﬁne what will happen during

the simulation. Firstly, the user will have to choose an

assay type in a dropdown menu (toxic, feeding, plate

tapping, touch) and then the window will show all the

options related to this experiment. For example, in a

touching experiment, the user will set some touches

with their pressure, their timing and the zone where

they will be done. As in the environment deﬁnition

window, some parameters will be set with numerical

values (time), with dropdown menus (different pres-

sure levels) and with the aid of the 3D window (zone).

Finally, in the 3D window, the starting point of

the simulation (obstacles, plate shape, position of the

worm, etc) are rendered using Three.js library. The

user will be able to explore the area with the mouse

and he will also use this window to deﬁne some as-

pects of the environment or the experiment, as stated

before.

Fig. 4 shows the ﬁrst prototype of the web page,

where the user has deﬁned an environment with some

obstacles and the simulation is running.

Once the experiment is completely deﬁned, the

same 3D window will be used to render the move-

ments of C. elegans. As in the experiment deﬁnition

tool, the user will be able to explore the area with the

mouse. At the moment, the prototype only renders the

skin of the animal, but in the future, the user will be

able to explore all the cells that compound the body.

In the future, the visualization of the behaviour of

the animal will be customizable. The user may want

to show only some parts of the experiment in the 3D

window, watch it in a slower velocity or stop it, etc.

The user will also see the tracing of the neurons

that he wanted to be traced. This tracing will have

VisualizationofaVirtualCaenorhabditiselegansinWebGL

167

been set in a different user interface. There will be

a list of neurons and their corresponding values in

a timeline (synchronized with the simulation). The

user will be able to remove or add neurons in the list

(among those that have been traced) and change the

tracing mode (voltage values, only spikes, etc.).

5 CONCLUSIONS AND FUTURE

WORK

In this paper we have shown our work to simulate the

locomotion of the worm Caenorhabditis elegans and

render it in a web browser without any plug-in in-

stallation. On the one hand, we created a physically-

based engine that simulates the locomotion of the ne-

matode. The body of the worm is compounded of

biphasic springs and different forces are applied to

them to obtain the locomotion: muscle forces, elas-

ticity, internal pressure and environmental forces.

On the other hand, we developed a web page

where the simulation is shown. The movements of the

muscles are transferred to a realistic 3D reproduction

of the animal using animation bones. The user can set

the parameters of the simulation (environment, posi-

tion of the worm, obstacles, etc.) and will be able

to deﬁne the parameters of the experiment (touching

moment, toxins, etc.). Once the simulation is deﬁned,

the web page will run it and will show the information

that the user requires.

Although Si elegans project is advancing fast,

there is still some work to do. Apart from making the

simulation of the locomotion more realistic, regarding

visualization, we identify two main lines of work. On

the one hand, we want to improve the transference of

the information from the physics engine to the web.

This way, we will be able to show all the parts of the

worm (muscles, neurons, organs) in the 3D environ-

ment and offer an innovative way to explore the work-

ing of the organism.

On the other hand, we aim to offer neuroscientists

a useful tool for the study of C. elegans neuronal ac-

tivity. For that, we have to develop a part in the web

page where all the required information will be shown

in a friendly and useful way.

ACKNOWLEDGEMENTS

The Si elegans project is funded by the 7th Frame-

work Programme (FP7) of the European Union un-

der FET Proactive, call ICT-2011.9.11: Neuro-Bio-

Inspired Systems (NBIS).

REFERENCES

Allard, J., Cotin, S., Faure, F., Bensoussan, P.-J., Poyer,

F., Duriez, C., Delingette, H., and L., G. (2007).

Sofa - an open source framework for medical simu-

lation. Medecine Meets Virtual Reality (MMVR15),

pages 1318.

Altun, Z. and Hall, D. (2009). Introduction. wormatlas.

Boyle, J. H., Berri, S., and Cohen, N. (2012). Gait mod-

ulation in c. elegans: An integrated neuromechani-

cal model. Frontiers in Computational Neuroscience,

6(10).

Bryden, A. and Cohen, N. (2008). Neural control of

caenorhabditis elegans forward locomotion: the role

of sensory feedback. Biol. Cybern., 98:339351.

Chalﬁe, M., Sulston, J., White, J., Southgate, E., J.N., T.,

and Brenner, S. (1985). The neural circuit for touch

sensitivity in caenorhabditis elegans. Journal of Neu-

roscience, 5:956964.

Geppetto (2014). http://live.geppetto.org.

Mailler, R., Avery, J., Graves, J., and N., W. (2010).

A biologically accurate 3d model of the locomotion

of caenorhabditis elegans. In Proceedings of the

2010 International Conference on Biosciences (BIO-

SCIENCESWORLD10), pages 8490.

Mujika, A., de Mauro, A., Robin, G., Epelde, G., and

Oyarzun, G. (2014). A physically-based simulation of

a caenorhabditis elegans. In 22nd International Con-

ference in Central Europe on Computer Graphics, Vi-

sualization and Computer Vision, pages 177184.

Niebur, E. and Erd

os, P. (1991). Theory of the locomotion

of nematodes. Biophysical Journal, 60:11321146.

OpenWorm (2014). http://www.openworm.org/.

OpenWormBrowser (2014). http://browser.openworm.org.

Palyanov, A., Khayrulin, S., Larson, S., and Dibert, A.

(2012). Towards a virtual c. elegans: A framework

for simulation and visualization of the neuromuscular

system in a 3d physical environment. Silico Biology,

11(3):137147.

Solenthaler, B. and Pajarola, R. (2009). Predictivecorrective

incompressible sph. ACM Trans. Graph., 28, 3.

Suzuki, M., Goto, T., Tsuji, T., and Ohtake, H. (2005). A

dynamic body model of the nematode c. elegans with

neural oscillators. Journal of Robotics and Mecha-

tronics, 17(3):318326.

Three.js (2014). http://threejs.org/.

Voegtlin, T. (2011). Clones : a closed-loop simulation

framework for body, muscles and neurons. BMC Neu-

rosci., 12.

WebGL (2014). http://www.khronos.org/webgl/.

NEUROTECHNIX2014-InternationalCongressonNeurotechnology,ElectronicsandInformatics

168