Proteotronics: Application to Human 17-40 and Bacteriorhodopsin

Receptors

Eleonora Alfinito

, Lino Reggiani

, Rosella Cataldo

, Giorgio De Nunzio

Maria Rachele Guascito

and Livia Giotta

Dipartimento di Ingegneria dell`Innovazione, Università del Salento, via Monteroni, Lecce, Italy

Dipartimento di Matematica e Fisica “Ennio de Giorgi”, Università del Salento, via Monteroni, Lecce, Italy

Dipartimento Di.S.Te.B.A., Università del Salento, via Monteroni, Lecce, Italy

Keywords: Protein Electrical Properties, Proteotronics.

Abstract: Human olfactory 17-40 and Bacteriorhodopsin are two protein receptors that received particular attention in

electronics, due to the possibility of implementing nano-biodevices able to detect odours and light and thus

useful for medical and green energy harvesting applications. Some recent experiments concerning the

electrical responses of these receptors are reviewed. Data are interpreted in the framework of a new science

exploiting the complexity in biology and biomedical engineering called proteotronics. In particular, the

single protein is modelled as an impedance network whose topological properties affect the electrical

response as measured by experiments.

1 FOREWORD

Recent advances in science and technology, such as

the development of techniques and devices for

health care and green and renewable energy, are the

successful products of the synergy among different

bailiwicks. As a matter of fact, cross-fertilization has

diffused a consolidated knowledge beyond the

boundaries of specific cliques, thus allowing the

birth of a more comprehensive methodological

approach which integrates and makes more powerful

chemical, computational, biological, engineering and

physical strategies (Alfinito, 2015a; De Nunzio,

2015; Guascito, 2011; Nagy, 2013).

Complexis 2016, being the first international

conference on complex information systems, offers

the relevant opportunity to introduce proteotronics, a

new emerging discipline aiming to propose and

develop innovative electronic devices, based on the

selective action of specific proteins. The word

originates from the combination of proteomics, the

science devoted to the large-scale study of proteins

structures and functions, and electronics, the science

devoted to the development of devices manipulating

electrical current to perform useful tasks.

2 INTRODUCTION

Proteins are the core of the cellular functions. They

are macro-assemblies of hundreds to thousands of

amino acids. Amino acids are taken by a set of about

20 elements and are structurally similar molecules,

which differ for the specific R-group. Proteins carry

out very different functions: they produce energy,

transform chemicals, build tissues, etc. Their

function is intimately connected to structure (Berg,

2002), which, in most cases, changes

(conformational change) when the protein performs

its activity. Therefore, a growing interest is devoted

to determine protein 3D structures in the different

phases of the protein activity. At present, the number

of classified structures of proteins in their native

state is quite large, while the structures of activated

proteins are quite a few. As a matter of fact, the

determination of the structure of proteins in their

active state is much more complex than in their

native state. This because structure and function are

strongly connected and the measurement of the

former may change the latter (Shrödinger, 1944).

Protein activation itself is a big challenge of

investigation, involving the mechanisms of internal

binding rearrangements (Kobilka, 2007) and the

modification of the protein free-energy landscape

(Alfinito, 2015b). These and other questions have

been recently investigated on a special family of

Alﬁnito, E., Reggiani, L., Cataldo, R., Nunzio, G., Guascito, M. and Giotta, L.

Proteotronics: Application to Human 17-40 and Bacteriorhodopsin Receptors.

In Proceedings of the 1st International Conference on Complex Information Systems (COMPLEXIS 2016), pages 32-38

ISBN: 978-989-758-181-6

proteins, the receptors, i.e. proteins able to capture

an external ligand (small proteins, molecules, light,

etc.), then converting this capture into a biochemical

signal. Recently, proteins have gained a primary role

in advancing electronics performances toward a

more friendly, green and renewable guise. The

cornerstone of this interest is the possibility to equip

electrical devices with the peculiar abilities of these

proteins in recognizing, selecting and capturing

specific ligand. These biochemical activities have a

natural portability to the electronic world since they

can be converted into electrical signals (Hou, 2007;

Benilova, 2008; Vidic, 2006; Jin, 2006; Ron, 2010;

Casuso, 2007). To take on this challenge it is

necessary to gain knowledge about the protein

features in vivo and in vitro, to achieve the

experience for producing electronic nanodevices, to

develop the ability for integrating organic and

inorganic components, to collect information

sufficient for producing a model of the physico-

chemical mechanisms underlying the device

response. The body of these studies has been

recently named proteotronics (Alfinito, 2015a), by

combining proteomics, the large-scale study of

proteins, and electronics. In particular, we already

showed some breaking results in protein electrical

measurements and the way these results are

interpreted into the framework of proteotronics. This

is carried out by representing the protein like a

complex network able to mimic the electrical

responses observed in experiments and to be

predictive for novel results.

In this paper we present the application of

proteotronics for two receptors, the human olfactory

receptor OR 17-40 and the bacteriorhodopsin. They

received wide experimental investigation for their

electrical properties as biological transducers when

used as active parts of electronic devices. We

selected those two cases as the best representative

among the experimental datasets on which the

theoretical model, drawn in the next section, has

been validated. The content is organized as follows:

Section 3 reports experimental and theoretical

results concerning the electrical responses of human

olfactory receptor OR 17-40, of high interest for the

realization of bioelectronic noses. Section 4 reports

experimental and theoretical results concerning the

light receptor bacteriorhodopsin, of great interest for

future applications in the field of green energy

production; finally, Section 5 sketches the main

conclusions.

3 HUMAN OR 17-40

3.1 Generalities and Experiments

Recent investigations have confirmed the possibility

of detecting the protein activity by using electrical

measurements (Vidic, 2006; Jin, 2006; Hou, 2007;

Benilova, 2008; Guascito, 2011; Nagy, 2014). All

the analyzed proteins (mainly olfactory and light

receptors) show a characteristic impedance response,

which strongly depends on the protein structure and

on the environmental conditions. In other terms,

measurements of the electrical responses are a

sensitive tool for detecting protein conformational

change (activation). The present case concerns the

activation of human olfactory receptor OR 17-40.

This protein shows a very good sensitivity to

helional and heptanal odorants (Levasseur, 2003;

Vidic, 2006; Benilova, 2008), although the results

deviate from a standard expectation and need for a

microscopic interpretation (Alfinito, 2011b).

Figure 1: Randles cell. Rs is the solution resistance and

Rp, Zw and CPE, the polarization resistance, the Warburg

impedance, and the constant phase element of the sample,

respectively.

Here we report experiments performed by using

the electrochemical impedance spectroscopy (EIS)

(Benilova, 2008).

The experimental set-up involves protein

receptors in their membrane fraction, anchored on

properly functionalized specific substrates and used

to detect the in vitro dose- response to their specific

ligands. In particular, EIS measurements were

performed under different concentrations of the

protein specific odorants and for each concentration

an impedance spectrum was recorded. The standard

procedure to interpret the impedance spectrum is to

give a simplified representation of the apparatus by

using a Randles cell, like that shown in Figure 1. In

particular, Rp is the element more sensitive to the

protein activation. Therefore, measurements of its

value with and without the odorant were performed.

At increasing values of the specific odorant

concentration, the polarization resistance shows a

peculiar bell-shaped behavior, centered at 10

-10

(Benilova, 2008). Furthermore, by using

CPE

R p

Proteotronics: Application to Human 17-40 and Bacteriorhodopsin Receptors

complementary techniques such as the differential

surface plasmon resonance (SPR) and the

differential bioluminescence response, also a second

peak was observed, at a higher concentration of the

odorant (around 10

-5

M) (Vidic, 2006). This kind of

response, that does not exhibit a saturation at the

highest concentrations, is unexpected.

3.2 Theory

The protein structure-function correlation is here

interpreted at the amino acid level, building a graph

whose nodes correspond to the amino acids. Each

node contains several data, like the amino acid

position, taken by the public data banks (Berman,

2002), and its electrical polarization in terms of a

specific dielectric constant (Alfinito, 2011a). These

data are used to assign the links between nodes that

are associated with an elemental impedance

responsible of charge transfer. The procedure to

construct the analogous network of the protein

requires two steps. In the first step, an interaction

radius, R

, is assigned. It determines the degree of

the graph nodes, because two nodes are connected

only if the physical distance between the

corresponding amino acids is less than R

(Figure

2).

Figure 2: Graphical representation of human OR 17-40 in

its native state. The network is obtained by using R

= 6 Å.

Furthermore, this kind of network preserves the

memory of the protein structure, i.e. it changes if the

protein 3D structure changes. In the second step, the

protein function is then introduced, by attributing to

the links the role of a specific physical interaction.

In the present case, it is an electromagnetic

interaction, that describes the response of the protein

to different electrical solicitations, thus the links

correspond to elementary impedances. In particular,

since we are interested in monitoring the impedance

variation due to the protein activation, we use the

impedance of a simple RC parallel circuit, like the

couple CPE-Rp in the Randles cell.

Finally, the elemental impedance between the i,j-

th nodes is:

i,j

-1

i,j

i.j 0

ρ +iεεω

(1)

where A

i,j

- l

i,j

/4), is the cross-sectional area

between two spheres of radius R

centered on the i-

th and j-th node, respectively; l

i,j

is the distance

between these centers, ρ is the resistivity, taken to be

the same for every amino-acid;

i= -1 is the

imaginary unit, ε

is the vacuum permittivity, ω is

the circular frequency of the applied voltage. The

relative dielectric constant of the couple of i-th and

j-th amino-acids, ε

i,j

, is expressed in terms of the

intrinsic polarizability of each amino acid. The

network is connected to an external bias by using

ideal contact on the first and last amino acid and

solved by using standard techniques. In particular,

analogously to the well known Hodgkin-Huxley

model, the problem statement consists in a set of

linear equations whose solution is performed by a

computational procedure, based on the Kirchhoff’s

laws. The network global impedance spectrum is

represented by a Nyquist plot for each configuration

of the protein 3D structure. The role of the

interaction radius, R

is still an open problem.

Recent investigations strongly suggest that it is

related to the level of protein activation (Kobilka,

2007; Alfinito, 2015b) and in the following section

we make use of this conjecture to interpret the

experiments.

Besides the electrical characterization, also a

preliminary description in terms of graph topology

was attempted. The molecule networks resulting

from setting the interaction radius to reasonable

values were found to be graphs with a small-world

structure (Albert, 2002). Small-world networks

emerge as intermediate configurations between the

limiting cases of regular lattices and completely

random graphs. Artificial generation of a small-

world network can be obtained by the Watts-

Strogatz model, in which a fraction p of the links of

a regular lattice is replaced with random links: of

course p is 0 for the regular lattice, and becomes 1

for a completely random network. In small-world

networks the distribution of node degrees is Poisson-

like, and the Average Path Length is short, like in

random networks, but the Clustering Coefficient is

quite high, like in regular lattices.

COMPLEXIS 2016 - 1st International Conference on Complex Information Systems

3.3 Results

As a preliminary test, we analyze the relative

variation of static impedance (resistance) for the

single protein in the presence/absence of the ligand.

Globally small differences are detectable and, in

agreement with experiments only two regions of R

values centred around 20 and 50 Å are of interest,

because the native state resistance exhibits a higher

value than the active state (Benilova, 2008). In

agreement with (Kobilka, 2007) we have postulateda

functional dependence of the odorant concentration

on the value of R

(Alfinito, 2011b).

In Figure 3 we report the relative resistance

variations, calculated with our model. Furthermore,

symbols describe the same quantity as observed in

experiments performed with helional and heptanal.

Figure 3: Relative resistance variation for human olfactory

receptor OR 17-40. Continuous line is the calculated

result, full circles ( empty squares) are the dose response

to heptanal (helional) by experimental data (Benilova,

2008), here given in terms of R

(Alfinito, 2011a).

The bump at lower value of R

, should be, in a

similar way, attributed to the second bump observed

by SPR. As we can see, the agreement is qualitative

and more close to the heptanal than to the helional

data.

Further improvement of the model are necessary

to quantitatively fit experimental outcomes (Alfinito,

2015b).

These results are confirmed by the analysis of the

impedance spectrum. The Nyquist plots, as

calculated by using the native and active state of OR

17-40 are reported in Figure 4. Continuous line

describes the receptor impedance spectrum in the

native state calculated by using R

= 50Å and the

dashed line to the active state with the same value of

. As a comparison, data obtained with protein

exposed to a concentration of 10

-10

M of heptanal,

normalized to the static impedance of the protein not

exposed to the odorant are also reported (dots)

(Alfinito, 2011b). The agreement is qualitatively

correct, i.e. the spectrum is represented by a simple

semicircle and the shrinking of the semicircle, when

going to the activated state, is also reproduced. As a

general outcome, the comparison between theory

and experiment is quite good and able to explain the

quite novel results concerning the protein response

to the specific ligands.

Figure 4: Normalized Nyquist plot of human OR 17-40.

Lines (continuous for native state, dashed for active state)

refer to calculated data, with R

=50Å. Dots refer to

experimental data for the protein with heptanal at a

concentration of 10

-10

M, crosses to the protein without

heptanal (Benilova, 2008).

Finally, concerning the small-world features of

the network, the node degree distribution was

calculated, giving a bell-shaped curve. The

calculation was done at two values of the interaction

radius, R

= 6Å and R

= 12Å, and the behaviour of

the frequency curve in the transition from the

smaller to the larger R

value was as expected, i.e.

the bell width increases while its mean value gets

larger like in the Poisson distribution, which is

considered a good approximation for random

networks (Figure 5).

The average path length L and the clusterization

coefficient C were also calculated, together with the

same values normalized to C

and L

(which are the

clusterization coefficient and the average path length

calculated for a regular lattice with equal number of

nodes N and equal mean degree, MD). The result

was a large C (far greater than O(N

-1

) which would

be typical of random graphs, and similar to C

) and

a relatively low average path length, which is

compatible with the hypothesis of a small-world

network with significant clusterization. This is

particularly true for the lower R

case (see Table 1).

Table 1: Network parameters for human OR 17-40.

N R

(Å) MD C C/C

L L/L

315 6 6.1 0.56 0.93 8.3 0.31

315 12 29.5 0.59 0.81 3.0 0.53

Proteotronics: Application to Human 17-40 and Bacteriorhodopsin Receptors

Figure 5: Degree distribution of human OR 17-40 (dashed

lines) and bacteriorhodopsin (continuous lines), obtained

by using R

= 6 Å and R

= 12 Å.

4 BACTERIORHODOPSIN

4.1 Generalities and Experiments

Another receptor of interest for applications is

bacteriorhodopsin (bR), a protein found in a

primeval organism, the Halobacterium salinarum,

specifically in a part of its cell membrane called the

purple membrane (PM). This membrane is a very

thin lipidic film of 5 nm, about the protein height,

and shows a quite stable structure.

Bacteriorhodopsin is able to convert the sun light

into an electrical potential across the host cell

membrane. In doing so, the conjugated dye, the

retinal, changes its structure, also inducing the

complete protein conformational change.

Recently, several experiment have analyzed bR,

in vitro, both in dark and light. The aim was to test

the possibility of recovering the activity of this

protein outside its natural environment and finally to

convert the activation due to light into an electrical

signal useful for technological applications.

In doing so, patches of PM were anchored on a

conductive substrate and connected to an external

circuit. The connection was made mainly with two

different techniques, i,e, by using an extended

transparent conductive contact (Jin 2006, Ron 2010)

or the tip of a c-AFM (Casuso, 2007;

Mukhopadhyay, 2014).

These investigations focused on the

measurement of the current-voltage (I-V)

characteristics in static conditions and revealed the

great stability of bacteriorhodopsin toward thermal,

mechanical and electrical stress (Casuso, 2007). This

protein shows a medium-gap conductivity in dark

which can be significantly enhanced by light. The

observed I-V characteristics are super-linear, and

this feature becomes more evident by increasing the

applied bias. Therefore, the charge transport across

the protein is mainly attributed to a tunneling

mechanism. Tunneling can be described like the

crossing of rectangular barriers at low bias (direct

tunneling), or the crossing of triangular barriers

(injection tunneling), at high bias. In the latter

regime, a growth of the protein current of about 5

orders of magnitude has been observed (Casuso,

2007). All these impressive features require a deep

investigation about the microscopic origin of the

responsible mechanisms. Furthermore, they have

arisen great expectations in technology, so inspiring,

for example, the realization of a Grätzel cell based

on it (Renugopalakrishnan, 2014).

It has been observed that the light protein

activation may produce an enhancement of the

current as large as the 100% (Jin, 2006). In the

perspective of a future use of this protein in solar

cell, this outcome is of extraordinary relevance.

4.2 Theory

The network protein analogue has been also used to

describe the observed I-V characteristics in bR and

other proteins (Alfinito, 2015a). In Figure 6, the plot

of the protein graph, drawn by using R

= 6 Å, is

reported. As a matter of fact, light induces a

conformational change that this model is able to

account for. Accordingly, a sequential tunneling

mechanism is introduced by using a stochastic

approach within a Monte Carlo solving technique.

Figure 6: Graphical representation of bacteriorhodopsin in

its native state. Network is obtained by using R

= 6 Å.

The non-linear shape of the I-V characteristic,

reported in (Jin, 2006; Casuso, 2007), is accounted

for by considering a tunneling probability including

direct and injection regimes, respectively, as in

(Alfinito, 2011a; Alfinito, 2014):

COMPLEXIS 2016 - 1st International Conference on Complex Information Systems

Φ≥

⎟

⎠

⎞

⎜

⎝

⎛

β−=

Φ<

⎟

⎠

⎞

⎜

⎝

⎛

−Φβ−=

eV ,

2eV

expP

eV ,

expP

(2)

where V

is the potential drop between the couple of

i-j amino acids,

=/2m

2lβ

⎟

⎠

⎞

⎜

⎝

⎛

, m the electron

effective mass. In this model, to tunnel an energy

barrier is equivalent to reduce the resistivity from its

maximum value down to ρ

(V):

(V) =

min

(3a)

in direct tunneling regime (rectangular barrier), and:

Φ≥

⎟

⎠

⎞

⎜

⎝

⎛

−ρ+

ρ=ρ

Φ<ρ=ρ

eV ,

)V(

eV ,)V(

minMAX

MAX

(3b)

in the injection tunneling regime (triangular barrier).

4.3 Results

The I-V characteristics of bR in dark and light have

been calculated by using the aforementioned

theoretical approach. In particular, the model

parameters have been tuned on the data given by

(Casuso, 2007), which cover the largest bias range.

In such a way, the experiments have been

reproduced with good accuracy (Alfinito, 2011a).

Data in light were given by (Jin, 2006) in the range

0-1 V. A fine agreement with these data has been

found by taking into account the multiple effects that

light produces on a sample of light receptors. In

particular, light irradiation transfers energy to the

single protein by means of two basic processes: (i)

the protein activation, specifically the change of its

3D structure consequent the ligand capture; and (ii)

the protein excitation, specifically the increasing of

its free-energy without conformational change

(Alfinito, 2015).

In the framework of our model, we introduce

process (i) by changing the 3D protein structure

input, while process (ii) is described by changing the

value of R

. Finally a sample of proteins irradiated

with light of appropriate wavelength experiences

both the processes, with some proteins activated and

some proteins excited. The percentage of

activated/excited proteins has been fitted by using a

Hill-like equation, and it is associated with a specific

value of R

(Alfinito, 2015). Therefore, following

this scheme, the current response of bR samples has

been reproduced by using a binary mixture of native

and active states:

(i) the sample in light corresponds to R

= 6.3 Å

with 96% of activated proteins, (ii) the sample in

dark corresponds to R

= 5.8 Å and 100% of

proteins in the native state. By using these

guidelines, we reproduce the photocurrent measured

in experiments. Figure 7 reports the photocurrent of

bR calculated within our model, in the same bias

range explored by (Casuso, 2007).

Figure 7: Photocurrent for the bacteriorhodopsin,

calculated as described in the text. In the inset, the same

data in the bias range 0-1 V and the experimental

outcomes (circles) (Jin, 2006).

Furthermore, the theoretical approach is able to

foresee photocurrent over a larger (a priori, arbitrary

large) bias range. At present only the bias range

described in the inset has been explored by

experiments (Jin, 2006).

Also for this protein we checked the network

small-world character. The node degree distribution

was again bell-shaped (not shown for the sake of

brevity) and very similar to what was found for OR

17-40 (see Fig.5). The (absolute and relative)

average path length and clusterization coefficient are

shown in Table 2.

Table 2: Network parameters for bacteriorhodopsin.

N Rc (Å) MD C C/C

L L/L

222 6 6.2 0.58 0.96 7.0 0.37

222 12 28.9 0.59 0.82 2.6 0.60

Again, the lower R

case is more easily framed

in a small-world context, having large C/C

and low

L/L

5 CONCLUSIONS

Proteotronics is an emergent branch of electronics,

Proteotronics: Application to Human 17-40 and Bacteriorhodopsin Receptors

able to describe the electrical properties of proteins.

The model is physically plausible and sufficiently

flexible to be tailored for describing different

experimental conditions. Here we have investigated

two specific case concerning human OR 17-40 and

bacteriorhodopsin protein receptors with promising

chances of being used in developing a new

generation of electronic devices. The results are also

of basic interest in advancing the present knowledge

on the microscopic mechanisms responsible of

protein functioning inside living cells.

REFERENCES

Albert, R. and Barabási, A.L. 2002. Statistical mechanics

of complex networks. Reviews of modern physics,

74(1), p.47.

Alfinito, E, Millithaler, J-F. and Reggiani, L. 2011.

Charge transport in purple membrane monolayers….

Physical Review E 83, no. 4 pp. 042902.

Alfinito, E., Millithaler, J.F., Reggiani, L., et al. 2011.

Human olfactory receptor 17-40 …. RSC Advances,

1(1), pp.123-127.

Alfinito, E and Reggiani, L. 2014 Opsin vs opsin: New

materials for biotechnological applications. Journal of

Applied Physics 116 (6), p. 064901.

Alfinito, E., Pousset, J.-F., and Reggiani, L. 2015.

Proteotronics: Development of Protein-Based

Electronics, Pan Stanford, Singapore.

Alfinito, E. and Reggiani, L. 2015. Mechanisms

responsible for the photocurrent in bacteriorhodopsin.

Physical Review E, 91(3), p.032702.

Benilova, I.V., Vidic, J.M., Pajot-Augy, et al. 2008.

Electrochemical study of human olfactory receptor OR

17–40…. Materials Science and Engineering: C,

28(5), pp.633-639.

Berg, J.M., Tymoczko, J.L, Stryer, L. 2002 Biochemistry,

W H Freeman, New York, V edition.

Berman, H.M., Westbrook, J., Feng, Z., et al. 2000. The

protein data bank. Nucleic acids research, 28(1),

pp.235-242.

Casuso, I., Fumagalli, L., Samitier, J., et al. 2007.

Nanoscale electrical conductivity of the purple

membrane monolayer. Physical Review E vol 76(4),

pp. 04191.

De Nunzio, G., Cataldo, R. and Carlà, A. 2015. Robust

Intensity Standardization …. Journal of digital

imaging, pp.1-11.

Guascito, M.R., Chirizzi, D., Malitesta, C. et al. 2011.

Mediator-free amperometric glucose biosensor ….

Analyst, 136(1), pp.164-173.

Hou, Y., Jaffrezic-Renault, N., Martelet, et al. 2007.A

novel detection strategy for odorant molecules…

Biosensors and Bioelectronics, 22(7), pp.1550-1555.

Jin, Y., Friedman, N., Sheves, M., He, T. and Cahen, D.

2006. Bacteriorhodopsin (bR) as an electronic

conduction medium: Current transport through bR-

containing monolayers. Proceedings of the National

Academy of Sciences, 103(23), pp.8601-8606.

Kobilka, B.K. and Deupi, X. 2007. Conformational

complexity of G-protein-coupled receptors. Trends in

pharmacological sciences, 28(8), pp.397-406.

Levasseur, G., Persuy, M.A., Grebert, D., et al. 2003.

Ligand-specific dose–response…. European Journal

of Biochemistry

, 270(13), pp.2905-2912.

Nagy, L., Magyar, M., Szabo,T., et al. 2013. Photosyntetic

Machineries in Nano-Systems. Current Proteins and

Peptide Science, I5 pp.363-373

Renugopalakrishnan, V., Barbiellini, B., King, C., et al.

2014. Engineering a Robust Photovoltaic Device with

Quantum Dots and Bacteriorhodopsin. J. Phys. Chem.

C, 118, pp.16710-16717

Ron, I., Sepunaru, L., Itzhakov, S., et al 2010. Proteins as

electronic materials…Journal of the American

Chemical Society, 132(12), pp.4131-4140.

Shrödinger, E. 1944 What is life? Cambridge University,

Cambridge.

Vidic, J.M., Grosclaude, J., Persuy, M.A., et al. 2006.

Quantitative assessment of olfactory receptors

activity… Lab on a Chip, 6(8), pp.1026-1032.

COMPLEXIS 2016 - 1st International Conference on Complex Information Systems