LANDSCAPING THE FRAMEWORK OF BIO RESEARCH PROJECT

Generation of the 3D Atlas for Drug Target Discovery

Byung-Cheol Kim

and Sunghoon Kim

1,2

Medicinal Bioconvergence Research Center, 8th ﬂoor, B-dong, Advanced Institutes of Convergence Technology, Iui-dong,

Yeongtong-gu, 443-270, Suwon, Gyeonggi-do, Republic of Korea

College of Pharmacy, Seoul National University, San 56-1, Shillim-dong, Kwanak-gu, 151-742, Seoul, Republic of Korea

Keywords:

Scientiﬁc information visualization, Project management, Drug target discovery.

Abstract:

As the size of pharmacological research organizations is getting bigger, it is ever more critical to grasp the

entire R&D scene as a vivid image, because a lack of such an image makes it hard to understand a project as

a whole and thus to make high-level decisions. We provide one prototype of such a kind so that all the re-

searchers of a project can intuitively share the current situation, recognize the bottlenecks, and collaborate with

one another to resolve the problems. Our approach is to exploit the faculty of human vision, especially spatial

perception and memory, by deﬁning the axes of information space to be commensurable and interpolatable,

making shapes of the process and data on those axes, and providing referential cues in the space.

1 INTRODUCTION

Over the past decade, companies and laboratories in

the pharmaceutical industry are getting bigger ever

and the mission to identify and validate a new drug

target involves more complicated, different but corre-

lated processes and data. Although an overwhelming

number of data are gushing out from laboratories on

a daily basis, stakeholders can hardly get the integral

information to make high-level decisions like whether

to increase or reduce the budget, the next-step R&D

direction, even whether or not to kill the project, etc.

In other words, a lack of the entire image of the situ-

ation could impede, defer or dismantle a huge target

discovery project which has been funded hundreds of

millions of dollars.

Since a human can only deal with just a few

chunks of symbolic information at the same time due

to the limitation of human memory, the only way to

simultaneously deliver a number of data is to make

them as images which can be recognized at once.

The problem is that, in many cases, the informa-

tion of scientiﬁc projects is heterogeneous and even

non-geometric. Naive visualization of such informa-

tion often leads to just another heavy and hardly-

perceptible set of data elements. The data should be

well organized in a geometrically informative way to

exploit the faculty of human vision optimized for ge-

ometric shapes. Therefore, the crux of the problem

is how to deﬁne the information space in which the

processes and data are visualized so that they can be

seen as a kind of geometrically meaningful shapes.

We provide one prototype visualization frame-

work, a 3-dimensional atlas for the drug target dis-

covery, by establishing the criteria for deﬁning such

axes of information space and referential cues.

2 RELATED WORK

Traditionally, biomolecular networks and interactions

have been visualized in two dimensions. Cytoscape

is a representative software to layout and query

biomolecular interaction networks extracted from

high-throughput expression data and other molecu-

lar states (Shannon et al., 2003). It is still popular

because of its ease of use to visualize a network of

many nodes and links. The Systems Biology Graphi-

cal Notation (SBGN) is a kind of 2-dimensional(2D)

circuit diagrams to represent networks of biochemical

interactions (Novre et al., 2009). The Biological Con-

nection Markup Language (BCML) is deﬁned to de-

scribe biological pathways based on the SBGN (Bel-

trame et al., 2011). KEGG, Kyoto Encyclopedia of

Genes and Genomes (http://www.genome.jp/kegg/),

is a well-known database including pathway data and

circuit-like diagrams. These approaches are very use-

ful to ﬁnd out speciﬁc mechanisms and related data.

356

Kim B. and Kim S..

LANDSCAPING THE FRAMEWORK OF BIO RESEARCH PROJECT - Generation of the 3D Atlas for Drug Target Discovery.

DOI: 10.5220/0003873103560359

In Proceedings of the International Conference on Bioinformatics Models, Methods and Algorithms (BIOINFORMATICS-2012), pages 356-359

ISBN: 978-989-8425-90-4

 2012 SCITEPRESS (Science and Technology Publications, Lda.)

Table 1: An example target inventory in a tabular form. There are too many data items to examine in order to comprehend

the status of the entire drug target inventory. In addition, the number of targets and their properties can rapidly grow even

to hundreds and thousands. (Quantiﬁcation of non-digit information should be deﬁned carefully in accordance with speciﬁc

categories of data).

No TargetID

Disease Mode of

Structure Assay Biomarker

Lead Clinical

Indication Action Availability Validation

1 AIMP2 NSCLC Yes(1.0) No(0) N/A 0.6 No(0) Yes(1.0)

2 CDH6 Renal Cancer Yes(1.0) 0.5 Yes(1.0) 0.1 0.9 0.5

3 AIMP3 Lymphoma 0.7 Yes(1.0) 0.3 No(0) 0.4 0.2

4 CD23 RA Yes(1.0) No(0) Yes(1.0) No(0) 0.1 No(0)

5 Gp96 Lupus Yes(1.0) No(0) 0.8 Yes(1.0) 0.9 Yes(1.0)

But they always provide all the detailed information

or the entire data at the same time, which cannot be

comprehended ﬁguratively when the number of data

becomes bigger than even hundreds.

Thus, for the large number of data, systems have

been developed beyond 2D planar visualization to

fully exploit the 3-dimensional(3D) human percep-

tual system. Arena3D was suggested as layered vi-

sualization of biological networks in 3D (Pavlopou-

los et al., 2008). The layers make a group of nodes

distinctive to the other groups. FORG3D provides a

force-directed 3D graph editing capability for visual-

ization of genome scale data (Jussi Paananen, 2009).

Nodes and their links are automatically positioned by

simulating them as a force-directed mesh grid. Inter-

actorium(Skyrails) is a software platform to visualize

protein-protein interaction and to navigate the interac-

tion link in 3D (Widjaja et al., 2009). It is based on a

scriptable command system so that the user can pro-

gram his/her own menus and visualization methods.

Systems and methods above provide 3D visualization

frameworks but still draw all the raw data which can

be too big to take as informative images.

Out of nodes and links, there has been another ap-

proach to provide a kind of prospect on the informa-

tion as a terrain map or landscape. A gene expression

map for C. elegans was created by drawing it as ter-

rain in 2.5D (xy-plane and z-elevation) (Kim et al.,

2001). GeneTerrain is a kind of 2.5D terrain map

which focuses on visualizing the difference map of

expression data (You et al., 2008). Recently, Pﬁzer’s

research centre developed a drug target landscape, a

visualization system to convey the overall meaning

of an integrated set (Campbell et al., 2010). It pro-

duces a concept of zone in which the plotted data are

grouped geometrically and meaningfully. This kind

of approach can convert data into ﬁguratively infor-

mative things. Thus we try to elaborate it further in

the next section by establishing some principles for

such visualization.

3 PRINCIPLES

There are two kinds of consideration to visualize non-

geometric information. One is how to draw the infor-

mation and the other is how to watch it. More speciﬁ-

cally on the former, it is about where to position each

data in space. That leads to the deﬁnition of the axes

of information space on which data are plotted to be

viewed with perceptual ease.

The axes of information space for perceptual visu-

alization should be

Commensurable. The data item visualized on the

same axis should be the same kind, conceptually at

least. For example, a mark pointing to an item on

the left side must be able to be said less or more ’...’

in a certain quality than one on the right side. The

bounded length or elongation of plotted items can ex-

press the level, state, and/or quality of the axial infor-

mation. Then, this might help the user’s recognition

of the information as a single factor instead of a mere

collection of items.

Interpolatable. Each item on the commensurable

axis must be able to be tightly coupled with its neigh-

bours to the extent that any point between two items

can be meaningfully drawn by mixing the values of

the two. For instance, a point value between two given

point values, p

and p

, can be positioned with the

mixing parameter, t, as follows:

p(t) = (1 −t)p

+t p

(0 ≤ t ≤ 1)

Consequently, the point, p(t), describes a line which

is a geometric object out of non-geometric data.

Therefore, the plotted data in the 2D or 3D in-

formation space based on such axes can be seen as a

shape, which is easy to remember and even get some

intuition about future R&D directions, because the

user can think of it not as a set of data but as a sort

of shape. For example, growing or shrinking of the

shape can convey the feeling of aliveness of a project.

Hence the user can consider how to rescue it by ob-

serving and analyzing the evolution of the shape. The

user can also see the evolving history by visualizing

LANDSCAPING THE FRAMEWORK OF BIO RESEARCH PROJECT - Generation of the 3D Atlas for Drug Target

Discovery

357

Figure 1: An example 3D atlas for identiﬁcation and validation of druggable targets on the three axes, i.e., targets, druggability,

and disease indication. Each data item is drawn as a cube which has non-geometric properties such as brightness, hue, and

transparency for the sort of targets, the kind of diseases, and the degree of progress, respectively.

the data over time.

The latter is about how to facilitate spatial percep-

tion. The essence of spatial perception comes from

movement, i.e., translation and rotation. One can feel

the position by translating and the orientation by ro-

tating. Perception of the position and orientation of

an object is generally relative to something else, more

formally, ﬁducial points or referential objects. Hence

navigating the information space without any refer-

ence would give the user a chance to fall into a pitfall

of lost-in-space. There can be two kinds of scheme to

provide spatial reference, intrinsic and extrinsic. The

intrinsic reference is a unique geometric and/or col-

ored parts of the entire shape. So the user can catch up

his or her relative position by remembering the (pat-

tern of) position of those unique parts. The extrinsic

reference is a kind of compass positioned in some cor-

ner of the space or the user’s screen. The user can in-

stantly recognize the current position and orientation

in the space. The former is good to apply when there

are few objects to inspect and they are simple enough

to embed additional unique shapes and/or colors. On

the other hand, the latter would be better if there are

too many objects to view in the space.

4 CASE STUDY: A TARGET

DISCOVERY PROCESS

We have been developing a 3D atlas to visualize the

target inventory. For instance, a target inventory can

be described in a tabular form such as Table 1. The

number of targets listed up will be grow rapidly and

there can be more properties like in vivo efﬁcacy of

lead, disease model, patent, and so on. All the items

in this tabular form, however, can hardly be retained

together in the human memory in an organized form.

Thus, in order to make them as images to be perceived

at a moment, we deﬁned three axes of information

space, targets (x-axis), druggability (y-axis), and in-

dication of disease (z-axis).

The x-axis is for targets, i.e., items of the same

x-coordinate indicate properties of the same drug tar-

get. Neighboring targets are not randomly placed but

arranged by their functional proximity. So the targets

on the x-axis evolve from a target niche of primary in-

terest to other targets that show functional proximity

to the initial targets. The y-axis represents the ma-

turity of each target with respect to druggability. In

other words, the longer the elongation on the y-axis

is, the more druggable the target becomes. The axial

information covers seven categories such as a target

ID, mode of action, assay/biomarker, animal model,

structure, availability of lead, and clinical validation.

The z-axis is for diseases for which a target can be a

BIOINFORMATICS 2012 - International Conference on Bioinformatics Models, Methods and Algorithms

358

new drug candidate. Then, a target’s indication of dis-

ease will marshal on the z-axis from diseases of initial

concern to other pathologically highly-connected dis-

eases.

Therefore, these axes create a unique space for the

evolving target inventory. Each data item is drawn as

a cube on those axes so that a set of cubes can be per-

ceived as a big cube being built up like Lego blocks.

The solidness and volume of the big cube can convey

the feeling of completeness and richness of validated

druggable targets. A cube has three non-geometric

properties, transparency, hue, and brightness which

represent the degree of progress, the kind of diseases,

and the sort of targets, respectively. These hue and

brightness can function as the referential cue when

translating and rotating the entire cube. Figure 1 de-

picts an example 3D atlas for drug target discovery.

5 CONCLUSIONS

The problem itself is simple. The entire image of the

R&D scene is needed and should be generated based

on real data. The constraint is that the image ought

to be geometrically informative to exploit human spa-

tial perception. Thus the properties of the information

axes are deﬁned so that a collection of data items can

appear as a geometric shape. To appreciate and re-

member the shape concretely, a referential cue is sug-

gested to be an anchor point in the information space

in which the user can often be confused or lost.

The suggested prototype has partly shown the

proof of concept for the drug target discovery process

in 3D space. But the information axes of the 3D atlas

should be elaborated to be more commensurable and

interpolatable. Finally, to more facilitate the user’s

spatial perception, motion guidance for inspecting the

3D atlas such as navigational paths along the patho-

logical connections will be developed.

ACKNOWLEDGEMENTS

This research was supported by Global Frontier Re-

search Grant NRF-M1AXA002-2010-0029785 of the

National Research Foundation funded by the Ministry

of Education, Science and Technology of Korea.

REFERENCES

Beltrame, L., Calura, E., Popovici, R. R., Rizzetto, L.,

Guedez, D. R., Donato, M., Romualdi, C., Draghici,

S., and Cavalieri, D. (2011). The biological connec-

tion markup language: a sbgn-compliant format for

visualization, ﬁltering and analysis of biological path-

ways. Bioinformatics, 27(15):2127–2133.

Campbell, S. J., Gaulton, A., Marshall, J., Bichko, D., Mar-

tin, S., Brouwer, C., and Harland, L. (2010). Visualiz-

ing the drug target landscape. Drug Discovery Today,

15(1/2):3–15.

Jussi Paananen, G. W. (2009). Forg3d: Force-directed 3d

graph editor for visualization of integrated genome

scale data. BMC Systems Biology, 3:26.

Kim, S. K., Lund, J., Kiraly, M., Duke, K., Jiang, M., Stu-

art, J. M., Eizinger, A., Wylie, B. N., and Davidson,

G. S. (2001). A gene expression map for caenorhab-

ditis elegans. Science, 293:2087–2092.

Novre, N. L., Hucka, M., Mi, H., Moodie, S., Schreiber,

F., Sorokin, A., Demir, E., Wegner, K., Aladjem,

M. I., Wimalaratne, S. M., Bergman, F. T., Gauges,

R., Ghazal, P., Kawaji, H., Li, L., Matsuoka, Y., Vill-

ger, A., Boyd, S. E., Calzone, L., Courtot, M., Do-

grusoz, U., Freeman, T. C., Funahashi, A., Ghosh, S.,

Jouraku, A., Kim, S., Kolpakov, F., Luna, A., Sahle,

S., Schmidt, E., Watterson, S., Wu, G., Goryanin,

I., Kell, D. B., Sander, C., Sauro, H., Snoep, J. L.,

Kohn, K., and Kitano, H. (2009). The systems biology

graphical notation. Nature Biotechnology, 27:735–

741.

Pavlopoulos, G. A., O’Donoghue, S. I., Satagopam, V. P.,

Soldatos, T. G., Paﬁlis, E., and Schneider, R. (2008).

Arena3d: visualization of biological networks in 3d.

BMC Systems Biology, 2:104.

Shannon, P., Markiel, A., Ozier, O., Baliga, N. S., Wang,

J. T., Ramage, D., Amin, N., Schwikowski, B., and

Ideker, T. (2003). Cytoscape: A software environment

for integrated models of biomolecular interaction net-

works. Genome Research, 13:2498–2504.

Widjaja, Y. Y., Pang, C. N. I., Li, S. S., Wilkins, M. R., and

Lambert, T. D. (2009). The interactorium: visualis-

ing proteins, complexes and interaction networks in a

virtual 3d cell. Proteomics, 9(23):5309–5315.

You, Q., Fang, S., and Chen, J. Y. (2008). Geneterrain:

visual exploration of differential gene expression pro-

ﬁles organized in native biomolecular interaction net-

works. Information Visualization, 9:1–12.

LANDSCAPING THE FRAMEWORK OF BIO RESEARCH PROJECT - Generation of the 3D Atlas for Drug Target

Discovery

359