A Novel Computer Vision Methodology for Intelligent Molecular

Modeling and Simulation

Belal Medhat

and Ahmed Shawish

2,3

Department of Computer Science, The British University in Egypt, Cairo, Egypt

Faculty of Computer Studies, Arab Open University, Headquarters, Kuwait

Ain Shams University, Cairo, Egypt

Keywords:

Molecular Modeling and Simulation, Computer Vision, Molecular Graphics, Molecular Structures.

Abstract:

Molecular modeling and simulation tools are used to study the structure of the molecules for the purpose

of understanding and creating a new generation of technology that works on the nano-scale. The current

techniques mainly focus on visualizing the molecule’s structure using many illustrative methods, while they

leave the knowledge extraction load on the user that should be aware of many complex sciences. Developing a

new innovative method in this perspective becomes crucial to support such fast development in such vital ﬁeld

of sciences. This paper represents a novel computer vision method for molecular modeling and simulation

that gives the computer the ability to see and understand the structure of molecules just like the human eyes,

and also the ability to analyze its structure without human intervention. The proposed approach is based on

using the computer’s memory as a digital representation of the real 3D-physical scaled model of the molecule,

and hence accommodates machine learning techniques for an automated analysis job. Moreover, a parallel

processing approach has been adopted to speed up the whole process. The realistic case study of a glucose

molecule reports the outstanding performance of the proposed approach to model and analyze its structure

without human intervention. The proposed methodology makes the developing of an automated molecular

expert system a one step away.

1 INTRODUCTION

Today the world turns its eyes on the technologies

and phenomena that happens at the Nanoscale, where

scientists are studying the cell and the cellular struc-

tures such as proteins, their structures, and their func-

tions (Friedrichs et al., 2009; Durrant and McCam-

mon, 2011; Soni et al., 2014). Scientists are learn-

ing lessons from nature. They are looking forward to

building molecular machines and robots using mod-

iﬁed proteins and other nano

-materials to do things

that were impossible in the past. Molecular model-

ing and simulation tools can help scientists in study-

ing and modifying the structure of a molecule by

doing the following: visualizing the molecule us-

ing computer graphics, simulating the motion of the

molecule under different forces and conditions (Daw-

son et al., 2016; Khatib et al., 2011; Durn-Riveroll

et al., 2016; Jallu et al., 2012), simulating the interac-

tion between the molecule and other molecules (Lin-

1 Nano-meter = 1x10

−9

meter

dert et al., 2013; Friedrichs et al., 2009), analyzing the

arrangement and geometrical shapes of the atoms in-

side the molecule to ﬁnd the critical points at which

the molecule’s structure will change.

The current molecular modeling and simulation

tools like Avogadro (Hanwell et al., 2012), VMD

(Humphrey et al., 1996), YASARA (Krieger and

Vriend, 2014), and RasMol tools (Potterton et al.,

2002) pay much attention on rendering the molecule

structure and leave the user to study the molecule by

himself. They left all the analysis, and knowledge

extraction effort to be done by the user who must

be an expert in molecular sciences, and theories to

take over these tasks. Frequently, the user is even

obligated to write a computer program to customize

these tools in order to do very simple jobs. This user-

dependent approach makes the current methods and

tools away of getting the full beneﬁts from the com-

puter sciences’ methods and techniques like machine

learning, computer vision, and artiﬁcial intelligence.

Enhancing the current molecular modeling and simu-

lation tools to overcome their critical limitations be-

Medhat, B. and Shawish, A.

A Novel Computer Vision Methodology for Intelligent Molecular Modeling and Simulation.

DOI: 10.5220/0006576100970104

In Proceedings of the 11th International Joint Conference on Biomedical Engineering Systems and Technologies (BIOSTEC 2018) - Volume 3: BIOINFORMATICS, pages 97-104

ISBN: 978-989-758-280-6

come crucial to support the fast development in such

vital ﬁeld of sciences. This will help the scientists to

design new molecules and nano-materials that can be

used in many applications.

This paper represents a novel computer vision

methodology for molecular modeling and simulation

which mimics the human eye’s vision and gives the

computer the ability to see and understand the molec-

ular structures. It will enable the user to see inner

parts of the molecule that may be hidden using the

current techniques. It also has the ability to analyze

the molecular structures and extract rich knowledge

from it without human intervention. Hence, develop-

ing molecular expert systems, chemical and physical

knowledge bases will become a one step away.

The proposed methodology’s approach is based on

using the computer’s memory(RAM) as a 3D-digital

representation media to model the physical molecu-

lar structure (i.e, atoms and bonds) using digits 0 and

1. Each bit represents a cube of 1 picomemeter

in the spatial space of the molecule. A parallel

processing approach has been adopted to efﬁciently

speed up the process of the whole method. This pa-

per reviews most of the current molecular software

tools like RasMol, PyMOL, VMD, Avogadro, GRO-

MACS, and Jmol to discuss all their pros and cons.

The extensive simulation studies conducted on a glu-

cose molecule report the outstanding capability of

the proposed method to extract knowledge from the

molecule structure and analyze it without human in-

tervention, contrarily to current human-dependent ap-

proaches.

The rest of the paper is organized as follows: Sec-

tion 2 provides a scientiﬁc background in molecu-

lar modeling and simulation. It also reviews related

work. Section 3 provides a detailed explanation of

the proposed methodology and the suggested paral-

lel architecture. Section 4 presents the conducted real

molecular case study and compare between the pro-

posed methodology versus the current methods. The

paper is then concluded in Section 5.

2 BACKGROUND

This section explains how the molecular structure is

discovered using X-ray. It then reviews the related

work in molecular modeling and simulation.

2.1 Scientiﬁc Background

As illustrated in ﬁgure 1, there is a cycle of steps to

discover the structure of any molecule. The ﬁrst step

in Figure 1: The cycle of discovering any molecules

Figure 1: The cycle of discovering any molecule’s structure.

structure to extract the target molecule from living or-

ganisms. Step two is to get a sufﬁcient amount of such

molecule and convert it to a crystal form. Step three

is to expose such crystal into an x-ray crystallogra-

phy device. This device shoots extensive x-ray beams

from different angles through the crystal and collects

the diffraction of such rays on a light-sensitive sheet.

The collected light intensities are then analyzed by a

computer program to reveal the position of each atom

in the molecule and its chemical type. Here, it is

worth to mention that there are other techniques that

do the same job like NMR, Mass Spectrometry, and

3D Electron Microscopy, where NMR, for example,

use the magnetic ﬁeld instead of the light diffraction.

Nevertheless, they all produce a ﬁle called MOL as

an abbreviation of the word molecule that reveals the

position of each atom in the molecule and its chemi-

cal type. Again, there are other types of ﬁles differ-

ent than the MOL like SDF, XYZ, and PDB chemical

ﬁles. The SDF and XYZ ﬁles reveal the same infor-

mation but in a different ﬁle structure, while PDB ﬁles

are used to describe proteins. Finally, once the MOL

ﬁle is created, any molecular viewer software takes

place to render the molecule is 3Dimensions on any

computers screen.

2.2 Related Work

In the early 90s, a real improvement in x-ray crystal-

lography and molecular imaging techniques has ap-

peared, since then researchers and scientists have tried

to build software tools to study different molecules’

structures. This software can be divided into two cat-

egories. The ﬁrst category includes molecular view-

ers, molecular editors and molecular designers (Han-

well et al., 2012; Humphrey et al., 1996; Sayle and

Milner-White, 1995; Potterton et al., 2002) all of

these software tools can visualize molecules. the sec-

ond category includes molecular dynamics simula-

tion tools that visualize the chemical reactions of the

molecules(Humphrey et al., 1996; Emsley and De-

breczeni, 2012; Dreher et al., 2013; Phillips et al.,

2005). Now, let’s review a group of molecular soft-

ware with their pros and cons.

The Avogadro tool (Hanwell et al., 2012) visual-

izes the molecules to the user on a computer’s screen

BIOINFORMATICS 2018 - 9th International Conference on Bioinformatics Models, Methods and Algorithms

using 3D computer graphics. It also enables the user

to choose an atom as the origin and rotate around it us-

ing the mouse buttons and the keyboard buttons. Nev-

ertheless, the user should always memorize the place

and the colors of the atom as well as its chemical types

in order to easily navigate through the molecule with-

out losing focus.

The VMD tool (Humphrey et al., 1996) visual-

izes the molecules to the user with different types of

graphical representations. It can do a lot of energy

calculations. It works through a special scripting lan-

guage, that should be used to initiate any complex job.

learning a new programming language to customize

the molecular graphics software to execute complex

and even simple commands require a lot of time and

effort.

The YASARA tool (Krieger and Vriend, 2014)

renders molecules using 3D graphics with less num-

ber of polygons and in less time, so even smartphones

can render large molecules faster with no hanging or

lagging. The YASARA tool has the advantage of

working on molecules anywhere and on any type of

computers from workstations to smartphones. how-

ever, this tool pays much attention to rendering and

rotating the molecules within a tedious workspace

without enough attention to real analysis and knowl-

edge extraction.

(Emsley and Debreczeni, 2012) designs drugs

using molecular graphics tool that can render the

molecules with different complex presentations.

Again, it is the user job to understand these presen-

tations that may take a long time based on his knowl-

edge.

3 THE PROPOSED S COMPUTER

VISION METHODOLOGY

3.1 The Main Idea

This section provides a detailed explanation of how

the 3D-physical model of the molecule is represented

in the computers memory, and how the knowledge is

extracted. It also explains the adoption of the paral-

lel architecture to speed up the process of the whole

methodology. The main idea of the proposed method-

ology is to build a digital model of 0 and 1 into the

computers memory in the form of a 3D-array of bits

that identically simulate the real 3D-physical struc-

ture of the molecule. Then, a computer vision al-

gorithm is proposed to help the computer to scan,

and analyze the constructed 3D-digital model with-

out the human intervention. Finally, a parallel pro-

cessing architecture has been adopted to speed up the

data processing of the 3D-array. Note that the pro-

posed approach is not a molecular viewing or ren-

dering approach, its an analysis methodology which

makes the computer able to understand and analyze

the molecules without human interference through the

proposed computer vision and knowledge extraction

methodology to be illustrated in the next subsections.

3.2 The Construction of the 3D

Molecular Representation

Here, we explain how we build the 3D-digital model

in form of a 3D-array of bits (i.e., 0 and 1) that

identically simulate the real physical structure of the

molecule. First, the proposed methodology reads the

input ﬁle that describes the structure of the molecule

in terms of its atoms that are inter-connected with spe-

ciﬁc chemical bonds

. Note that the input ﬁle de-

scribes any given molecule by listing its atoms and

their spatial distribution in a 3D-dimension. The

current molecular modeling and simulation software

tools render this input ﬁle using OpenGL library and

use any traditional graphical tools to visualize its

structure. In our approach, the same process is done

but on a 3D-array of bits. As illustrated in ﬁgure

2, each atom in this molecule is created in the form

of a sphere of bits of 1, where each bit represent

1Pico − meter

3 3

, which is the basic measuring unit

used to represent an atom. The rest of the bits in the

3D-array will be of value 0 to represent the empty

space between the atoms. Note that even the size of

each atom based on its chemical type is identically

reﬂected inside the new representation.

The result of this process will produce an identi-

cal digital model of the original molecule in the com-

puter’s memory (RAM) as illustrated in Figure 3. The

size of this array can vary depending on the original

size of the molecule in Pico-meter unit.

A library has been developed to hold the unique

features of all the organic atoms

, which help the

computer to recognize each atom later using the pro-

posed computer vision algorithm. To sum up, we can

now claim that this 3D-presentation reﬂects all major

information about the molecule’s structure such as its

atoms spatial distribution, size, and the internal dis-

tance between atoms.

The creation of this ﬁle was explained in section 2, and

it is the same ﬁle used as input for any of the existing tools

1 Pico-meter = 1x10

−12

meter

The organic atoms are: Oxygen, Nitrogen, Hydro-

gen, Sulphur, Phosphorus, Calcium, Magnesium, Potas-

sium, Chlorine, Sodium.

A Novel Computer Vision Methodology for Intelligent Molecular Modeling and Simulation

Figure 2: The atom in nature and as presented in 3D-array.

Figure 3: Building the whole molecule inside the 3D-array

of bits.

3.3 The Proposed 3D Computer Vision

Algorithm

We should remember that the generated 3D-array in

the previous step is a collection of zeros and ones that

are still in need for interpretation to extract valuable

information that can be later transformed to knowl-

edge using machine learning techniques. The pro-

posed computer vision algorithm explained in this

subsection will take over this task. Here, the computer

vision methodology is composed of two phases, the

ﬁrst phase is scanning the 3D-array, and the second

phase is extracting the knowledge from the scanned

data.

This algorithm takes either the whole molecule’s

3D-array as input or even a smaller part of it (i.e., a

sub 3D-array of bits) to understand and reveal all nec-

essary information about the examined area. It starts

from a speciﬁc atom in the molecule or even around it

within a given space. As illustrated in ﬁgure 4, the al-

gorithm starts by reading the given 3D-array of bits by

examining the 2D-array in XY plane and then move to

the next 2D-array in the Z-direction. It keeps record-

ing the countered spheres in a list that we call the

”Hit-List” and recognize each atom through its diam-

eter

. Note that the chemical bonds which exist be-

tween atoms are represented in a separate data struc-

ture which links each chemical bond with its end-

points atoms that are participating in the bond. This

process can be done starting from a speciﬁc origin

in a certain direction inside a volume with a given

depth. The scan works as a transformation and pro-

Scientiﬁc fact there are no two different organic atoms

with the same diameter

Figure 4: Two types of scan, type A scanning a cube around

an atom, and type B scanning everything in a cube which the

chosen atom lies in the center inside the cube being scanned.

jection from 3D to a 2D-presentation which produce a

stream of 2D-images of atoms inside the scanned vol-

ume space. The geometric arrangement of the atoms

inside the resultant 2D-images such as angles between

bonds’ axes and distances between atoms are stored

in the Hit-List. The scanning process reveals also the

geometric shapes formed by the atoms and the bond’s

axes. Note that the produced Hit-List will be the ba-

sic input for the next knowledge extraction phase. The

steps of the scanning algorithm are illustrated in Fig-

ure 5: Algorithm 1: Scanning the 3D-array.

Once the 3D-array is scanned, the computer will

be able to study the relationship between the atoms

using molecular geometry functions and extract the

following knowledge:

• The distances between the atoms.

• The bonds that exist between the atoms.

• The arrangements of the atoms.

• The electrical charges of each atom.

• The volume of the molecule.

• The volume of empty space inside the molecule.

• The distribution of the atoms’ density.

• The distribution of atoms’ weights.

• The geometric shapes that are formed by the

atoms and the bond between them.

• Empty volumes of space inside the molecule and

between its atoms.

• Recognition of the molecule’s surfaces and the

atoms that compose its surfaces in all directions.

• The dimension of the cuboid that encloses the

molecule.

As a matter of fact, the ﬁrst three extracted points

can be mathematically computed from the MOL ﬁle

directly, however, they are still easily calculated from

the new presentation and should be reported as part of

the extracted knowledge. Nevertheless, the remaining

BIOINFORMATICS 2018 - 9th International Conference on Bioinformatics Models, Methods and Algorithms

100

Figure 5: Algorithm 1: Scanning the 3D-array.

points cannot be extracted by any other method, like

computing the molecules outer surface and the geo-

metric shapes of the space between atoms. This is be-

cause of the interleaving between the atoms surfaces

inside the molecule. The proposed methodology uses

a 3D-graphics library to display the highlighted parts

of the molecule which are reﬂected in a comprehen-

sive report that reveals the extracted knowledge from

such structure. The proposed methodology main ob-

ject is not to simply render the molecule in a 3D as

same as the current tools, it is designed to analyze

the molecule and extract knowledge without any hu-

man intervention. It also enables us to use machine

learning techniques directly since the computer be-

comes now able to understand the internal structure of

the molecule. From the computational point of view,

we should highlight the processing time of the model

heavily depends on the size of the 3D-array. This is

why a parallel architecture has been adopted to over-

come this prospective limitation as explained in the

next subsection.

Figure 6: Distributing the sub cubes among the cores of a

CPU.

3.4 Parallel Processing

The 3D-array can be divided into sub 3d-arrays with-

out losing the spatial arrangements of atoms in the

molecule, and hence a strong potential for applying

the single instruction multiple data paradigm (SIMD)

on these sub 3d-arrays exists in order to accelerate the

scanning of the molecule’s representation by harness-

ing the underlying multi-core hardware. Note that the

time consumed by the CPU to build the 3D-digital

representation represents only 1% out of the total run-

time

, so the CPU can build and swap between the

cubes very fast.

The proposed algorithm uses the OpenMP C/C++

library for parallel processing of the molecule’s sub

3d-arrays. The proposed parallel implementation di-

vides the molecule’s cube into sub 3d-arrays and as-

signs each of them to one processing core or thread.

The previously mentioned scanning part is then im-

plemented on all the sub 3d-arrays simultaneously, as

illustrated in ﬁgure 6. After a careful study of both

scanning and knowledge extraction phases, we found

that a parallel processing architecture can be applied

in the scanning phase only while it is not suitable yet

for the knowledge extraction one due to the inter-

dependency between sub 3D-arrays at their bound-

aries. Therefore, all the information that has been col-

lected from the parallel scan are then collected cen-

trally in the Hit-List for the next knowledge extraction

phase.

4 EXPERIMENTAL CASE STUDY

AND COMPARISON

This section discusses the results of the extensive

simulation studies conducted on a realistic glucose

molecule using the proposed methodology.

4.1 Simulation Setting

The proposed methodology has been implemented

on a Linux operating system using C programming

This fact is deduced from a realistic experimental study

A Novel Computer Vision Methodology for Intelligent Molecular Modeling and Simulation

101

Figure 7: Performance and memory consumption measure-

ments per different number of threads.

language, OpenGL graphics library for checking the

results by rendering the molecules using 3D graph-

ics, and OpenMP for multi-core programming. The

studies were implemented on Intel Core i7-3612QM

with 6 mega cache CPU and 8 threads each thread

works at processing frequency range from 2.10GHz

to 3.10GHz and Intel Core i7-4790 with 8 mega cache

CPU and 8 threads each thread works at processing

frequency range from 3.60 GHz to 4.00 GHz.

The source code is divided into following four

main modules:

• The ﬁrst module is responsible for reading the

chemical ﬁles.

• The second module is responsible for constructing

the 3D-array of bits in the memory and building

the molecule’s atoms inside it.

• The third module is responsible for executing the

parallel scan of the molecule’s sub 3D-arrays.

• The fourth module is responsible for extracting

the knowledge and visualizing the results to the

user on the screen using 3D graphics.

4.2 Case Study

Simulation studies have been conducted on different

3D-arrays sizes distributed among a different number

of cores to assess the parallelism effect of the pro-

posed methodology and to measure the gain of speed-

ing up the scanning phase. As illustrated in ﬁgure

7, in the case of using 1 thread (serial program) the

run-time took 32 seconds and the memory consumed

is 0.2 Gigabytes. In case of 4 threads (Parallel pro-

gram) the run-time decreases by 72% and the mem-

ory consumption remain the same, after increasing the

number of threads to 8 the run-time decreases by 18%

again and the memory consumed increased by 50%.

Finally, in case of 16 threads the run-time increases

up by 2% and the memory consumption increased by

Figure 8: The glucose molecule in XY, XZ, and YZ planes.

Figure 9: The resultant report of the proposed tool.

33.3%. It is worth to note that the more threads we

apply on the CPU, the more the run-time and memory

consumed increase because allocating memory for

threads, creating and destroying threads costs over-

head processing run-time and memory. Based on the

previous run-time degradation, we preferred to imple-

ment the following case study using 8 threads in order

to get the best performance with the least run-time.

Once the 3D-array is scanned and the Hit-list is

created, the proposed methodology displays the glu-

cose molecule in 3D with six carbon atoms colored

in dark grey, six oxygen atoms colored in red, and

twelve hydrogen atoms colored in light grey, as illus-

trated in ﬁgure 8. The ﬁgure displays only 3 shots

for the molecule from the XY, XZ, and YZ planes

for simplicity as a printed version. The proposed

methodology also displays a comprehensive report

that reveals the knowledge extracted from the glucose

molecule As illustrated in ﬁgure 9. The report covers

the extracted information previously mentioned be-

fore in section 3.3.

To show the capability of the proposed method-

ology, we conducted a more sophisticated experi-

ment by choosing a speciﬁc atom inside the glucose

molecule and study only the area behind it for the pur-

pose of going into a deeper level of understanding of

each atom in the molecule. The proposed method-

ology also displays a comprehensive report that re-

veals the knowledge extracted as illustrated in ﬁgure

10. The report covers the following additional infor-

mation inside the scanned space:

BIOINFORMATICS 2018 - 9th International Conference on Bioinformatics Models, Methods and Algorithms

102

Figure 10: The geometric analysis of the proposed tool.

Table 1: Comparison between the available tools and the

proposed methodology.

Comparison criteria Available

tools

Proposed

method-

ology

User dependent Yes No

knowledge extraction No Yes

3D Rendering Yes Yes

Navigate inside the 3D

model

Yes Yes

View geometric patterns

inside the molecule

No Yes

Recognition of the

molecule surface

No Yes

The volume of empty

space inside the molecule

No Yes

The distribution of the

atoms density of the

molecule

No Yes

The distribution of atoms

weight of the molecule

No Yes

• The distances between the chosen atom and the

atoms behind it.

• The chemical bonds between the chosen atom and

the atoms behind it.

• 3D-images of the atoms spatial distribution.

• 3D-geometric shapes of the spaces between

atoms.

Next, the glucose molecule is examined using the

current tools like Avogadro and PyMol. As illustrated

in ﬁgures 11 and 12, the Avogadro and PyMol tools

render the glucose efﬁciently on the screen and wait

for the user to choose one atom as the center of rota-

tion. The user will rotate the glucose molecule using

the mouse and try to recognize the chemical types of

the atoms around the origin atom through their colors

Figure 11: Studying the glucose molecule using Avogadro

tool.

Figure 12: Studying the glucose molecule using PyMol

tool.

using his eyes. The user will try to study the rela-

tionships between the origin atom and atoms around

it using his knowledge in molecular sciences. The

user may have to spend some time in writing a small

script in a special programming language or click-

ing on menus and buttons in order to customize the

molecular viewer.

To sum up, we can easily note the outstanding ex-

tracted knowledge using the proposed methodology

in comparison with the currently available tools that

mainly depend on the user, as summarized in Table 1.

The proposed solution promises to open a new area of

molecular sciences and will signiﬁcantly enhance the

development in this crucial ﬁeld.

5 CONCLUSION

This paper represents a novel computer vision

methodology for molecular modeling and simulation

that gives the computer the ability to see, understand,

and analyze the molecular structures by itself with-

out human intervention. Its main idea was based on

using the computer’s memory (RAM) as a 3D-digital

representation of the molecule’s structure. A new al-

gorithm was developed to help the computer to see the

new representation, and extract the knowledge about

the vital aspects inside the molecule using a parallel

architecture to speed up the data processing. This pa-

per reviews most of the current molecular software

A Novel Computer Vision Methodology for Intelligent Molecular Modeling and Simulation

103

tools like RasMol, PyMOL, VMD, Avogadro, GRO-

MACS, and Jmol to discuss all their pros and cons.

The extracted knowledge reports the outstanding ca-

pabilities of the proposed methodology in comparison

with the current tools.

REFERENCES

Dawson, W. K., Maciejczyk, M., Jankowska, E. J., and Bu-

jnicki, J. M. (2016). Coarse-grained modeling of rna 3d

structure. Methods, 103(Supplement C):138 – 156. Ad-

vances in RNA Structure Determination.

Dreher, M., Piuzzi, M., Turki, A., Chavent, M., Baaden,

M., Frey, N., Limet, S., Rafﬁn, B., and Robert, S. (2013).

Interactive molecular dynamics: Scaling up to large sys-

tems. Procedia Computer Science, 18(Supplement C):20

– 29. 2013 International Conference on Computational

Science.

Durrant, J. D. and McCammon, J. A. (2011). Molecular

dynamics simulations and drug discovery. BMC Biology,

9(1):71.

Durn-Riveroll, L. M. et al. (2016). Docking simulation of

the binding interactions of saxitoxin analogs produced by

the marine dinoﬂagellate gymnodinium catenatum to the

voltage-gated sodium channel nav1.4. Toxins. Available

from: ”http://www.mdpi.com/2072-6651/8/5/129”. Acc-

sessed: [13 MAY 2016].

Emsley, P. and Debreczeni, J.

E. (2012). The Use of Molec-

ular Graphics in Structure-Based Drug Design, pages

143–159. Humana Press, Totowa, NJ.

Friedrichs, M. S., Eastman, P., Vaidyanathan, V., Houston,

M., Legrand, S., Beberg, A. L., Ensign, D. L., Bruns,

C. M., and Pande, V. S. (2009). Accelerating molecular

dynamic simulation on graphics processing units. Jour-

nal of Computational Chemistry, 30(6):864–872.

Hanwell, M. D., Curtis, D. E., Lonie, D. C., Vandermeersch,

T., Zurek, E., and Hutchison, G. R. (2012). Avogadro:

an advanced semantic chemical editor, visualization, and

analysis platform. Journal of Cheminformatics, 4(1):17.

Humphrey, W., Dalke, A., and Schulten, K. (1996). Vmd:

Visual molecular dynamics. Journal of Molecular

Graphics, 14(1):33 – 38.

Jallu, V., Poulain, P., Fuchs, P. F. J., Kaplan, C., and

de Brevern, A. G. (2012). Modeling and molecular dy-

namics of hpa-1a and -1b polymorphisms: Effects on the

structure of the 3 subunit of the iib3 integrin. PLOS ONE,

7(11):1–10.

Khatib, F. et al. (2011). Crystal structure of a monomeric

retroviral protease solved by protein folding game play-

ers. nature structural and molecular biology, 18:1175–

1177.

Krieger, E. and Vriend, G. (2014). Yasara viewmolecu-

lar graphics for all devicesfrom smartphones to worksta-

tions. Bioinformatics, 30(20):2981–2982.

Lindert, S., Bucher, D., Eastman, P., Pande, V., and Mc-

Cammon, J. A. (2013). Accelerated molecular dynamics

simulations with the amoeba polarizable force ﬁeld on

graphics processing units. Journal of Chemical Theory

and Computation, 9(11):4684–4691. PMID: 24634618.

Phillips, J. C., Braun, R., Wang, W., Gumbart, J., Tajkhor-

shid, E., Villa, E., Chipot, C., Skeel, R. D., Kal, L.,

and Schulten, K. (2005). Scalable molecular dynam-

ics with namd. Journal of Computational Chemistry,

26(16):1781–1802.

Potterton, E., McNicholas, S., Krissinel, E., Cowtan, K.,

and Noble, M. (2002). The CCP4 molecular-graphics

project. Acta Crystallographica Section D, 58(11):1955–

1957.

Sayle, R. A. and Milner-White, E. (1995). Rasmol:

biomolecular graphics for all. Trends in Biochemical Sci-

ences, 20(9):374 – 376.

Soni, S., Tyagi, C., Grover, A., and Goswami, S. K. (2014).

Molecular modeling and molecular dynamics simula-

tions based structural analysis of the sg2na protein vari-

ants. BMC Research Notes, 7(1):446.

BIOINFORMATICS 2018 - 9th International Conference on Bioinformatics Models, Methods and Algorithms

104