Human-Centered Design for the Efﬁcient Management of Smart

Genomic Information

Alberto Garc

ıa S.

, Mireia Costa

, Ana Le

, Jose F. Reyes

and Oscar Pastor

Pros Research Center, VRAIN Research Institute, Polytechnic University of Valencia, Valencia, Spain

Keywords:

Genomics, Human-Centered Design, Human-Computer Interaction, GenomIUm.

Abstract:

Genomics is a massive and complex domain that requires great efforts to extract valuable knowledge. Due

to the reduction in sequencing costs and the advent of Next Generation Sequencing, the amount of publicly

available genomics data has increased notably. These data are complex and heterogeneous, which makes

the development of intuitive and usable tools critical. However, bioinformatics tools have been developed

without oncsidering usability and User Interface design. As a result, there are relevant usability problems that

complicate the work of bioinformaticians. Human-Centered Design consists of a design approach that grounds

the User Interface design process on the needs and desires of users and can be a suitable solution to improve

the usability of new genomics tools. This work shows how intuitive and usable bioinformatics tools can be

produced using HCD principles.

1 INTRODUCTION

Genomics is a complex interdisciplinary ﬁeld that has

revolutionized how we understand medicine, disease

prevention, and treatment. In recent decades, the

amount of publicly available genomics data has in-

creased dramatically, outpacing by far our ability to

interpret it. The scientiﬁc community has made great

efforts to address this situation and developed thou-

sands of bioinformatics tools to analyze and interpret

genomics data. However, a number of these tools

have been developed without considering how their

design impacts user experience and knowledge ex-

traction, and there is a growing concern that current

approaches are insufﬁcient to deliver intuitive and us-

able User Interfaces (UIs) (Pavelin et al., 2012).

The UIs of these tools lack the quality and usabil-

ity of other tools that people come across in their daily

lives, such as word processing or spreadsheets (Java-

hery et al., 2004). Javahery et al. highlighted that the

design of bioinformatics tools is more complex and

less intuitive when compared to more general-purpose

UIs (Javahery et al., 2004).

Consequently, bioinformatics tools poorly repre-

sent the underlying concepts, and their interaction

https://orcid.org/0000-0001-5910-4363

https://orcid.org/0000-0002-8614-0914

https://orcid.org/0000-0003-3516-8893

https://orcid.org/0000-0002-9598-1301

https://orcid.org/0000-0002-1320-8471

mechanisms are challenging to understand. This leads

to usability issues that signiﬁcantly impact the efﬁ-

ciency and effectiveness of bioinformatics daily work

(Bolchini et al., 2009), who will struggle to ﬁnd valu-

able information for their research (Carpenter et al.,

2012). Carpenter et al. suggest that, in order to ﬁnd

valuable information more efﬁciently and increase the

adoption of bioinformatics tools, usability should be

a more important goal (Carpenter et al., 2012).

To produce usable tools in complex domains such

as bioinformatics, getting a precise understanding of

how users work is crucial. Correctly understanding

domain-speciﬁc tasks and the particularities of their

working context allows developers to improve UI de-

sign, making them more usable and efﬁcient (Svanæs

et al., 2008). Having the user as the central source

of information to design and develop UIs has multi-

ple beneﬁts (Pavelin et al., 2012). For instance, users

will be more likely to use a tool if they have guided

its design process, and user-friendly access to the data

potentially increases users’ scientiﬁc discoveries.

Although usability has been frequently ignored in

bioinformatics, it is a fundamental dimension for gen-

erating easy-to-use UIs. A well-designed UI allows

users to perform their tasks efﬁciently and facilitates

achieving their goals (Rimmer, 2004). As we have

mentioned before, improving the usability of bioin-

formatics tools requires focusing more on UI design

and user needs during their design process. To this

aim, Human-Centered Design (HCD) can be a help-

ful solution as it designs UIs focusing on user needs

García S., A., Costa, M., León, A., Reyes, J. and Pastor, O.

Human-Centered Design for the Efﬁcient Management of Smart Genomic Information.

DOI: 10.5220/0011635800003464

In Proceedings of the 18th International Conference on Evaluation of Novel Approaches to Software Engineering (ENASE 2023), pages 15-26

ISBN: 978-989-758-647-7; ISSN: 2184-4895

 2023 by SCITEPRESS – Science and Technology Publications, Lda. Under CC license (CC BY-NC-ND 4.0)

and desires, considering them during the entire design

process (Chilana et al., 2010).

HCD has been applied successfully in multiple

domains such as health services (Dopp et al., 2019),

education (dos Santos et al., 2019), or elder health

care (Bradwell et al., 2019), yet its use in genomics is

limited (Bolchini et al., 2009). Several authors have

studied the reasons causing this situation (Jaspers,

2009; Chilana et al., 2010; Pavelin et al., 2012;

de Matos et al., 2013). Their ﬁndings can be sum-

marized in the following points:

• Bioinformatics experts have historically relied on

command-line tools, and using HCD requires a

“cultural shift”.

• Bioinformatics data is highly complex and inter-

connected, and additional technical and scalabil-

ity constraints have to be considered. Besides, it is

a continuously evolving subject whose rules usu-

ally have plenty of exceptions.

• Using HCD techniques generates an initial delay

in the design process, and measuring the impact

of applying these techniques is a too fuzzy activ-

ity. HCD techniques improve scientiﬁc discovery

processes, but “discovery” is an intangible metric

and, therefore, difﬁcult to measure.

• The prior knowledge needed to carry out HCD

techniques in this domain adequately (human-

computer interaction, bioinformatics, and com-

puting) creates a gap between domain users and

developers.

• The usability validation, crucial to provide suc-

cessful solutions, needs to be carried out by

skilled UI designers, which is not always possi-

ble.

• The most valued aspect of a tool is its novelty

rather than its associated HCD work, lessening

down usability and UI aspects.

Over the years, the use of HCD has slowly in-

creased, but its adoption is far from commonplace.

Some examples of tools developed following an HCD

approach are the following:

• Sutcliffe et al. developed ADVISES, a scenario-

based visualization tool to support epidemiologi-

cal research (Sutcliffe et al., 2010).

• Valentin et al. redesigned the EB-eye search ser-

vice using prototyping and storyboarding tech-

niques to analyze user tasks, and their domain

mental model (Valentin et al., 2010).

• Missier et al. developed Taverna, a tool to com-

pose and enact workﬂows for the bioinformatics

community (Missier et al., 2010).

• Rutherford et al. investigated how bioinfor-

maticians interact with existing tools to explore

large DNA sequences to improve their usability

(Rutherford et al., 2010).

• De Matos et al. applied HCD methods to cre-

ate the Enzyme Portal webpage (de Matos et al.,

2013).

• Ko et al. developed Closha, a workﬂow manage-

ment system oriented to the analysis of massive

genomics data (Ko et al., 2018).

• Recchia et al. designed a genetic report for cystic

ﬁbrosis patients. (Recchia et al., 2021).

• Cutting et al. designed a panel report aimed at

easing clinicians’ work (Cutting et al., 2016).

To solve the existing usability problems, bioin-

formatics tools need to focus more on usability and

ease of use through high-quality UIs. However, the

use of HCD is far from typical. In this work, we

report how we followed a methodological HCD ap-

proach to design and develop an intuitive and easy-

to-use bioinformatics tool called Sibila. Sibila is a

conceptual model-based web platform for identifying

relevant DNA variations associated with genetic dis-

eases. We also validate the usability of Sibila with do-

main experts and discuss the beneﬁts and drawbacks

of following such an HCD approach.

2 METHODS

This section describes the methods used to design and

develop Sibila, detailing how we followed an HCD

approach. The development process of Sibila relied

on ﬁve artifacts:

• ISO 9241-210:2019 (Standardization, 2019):

(“Ergonomics of human-system interaction —

Part 210: Human-centred design for interactive

systems”) deﬁnes a set of standardized principles

for applying HCD.

• The Conceptual Schema of the Genome (CSG)

(Garc

ıa S. et al., 2021b): is a model that rep-

resents human genomics and its inner workings

from a holistic perspective, integrating several di-

mensions. It is the ontological framework used to

understand the working domain and communicate

with domain experts effectively. This schema also

guided the development of Sibila.

• The Concur Task Tree (CTT) Notation (Pa-

tern

o, 2003): is a formal abstraction that links

user interfaces with the task necessary to accom-

ENASE 2023 - 18th International Conference on Evaluation of Novel Approaches to Software Engineering

Table 1: The list of activities carried out during each phase.

Phase Input Activity Output

Speciﬁcation of

Context of use

— (1) Focus group Users’ context of use.

CSG (2) Conceptual view

generation

A conceptual view from the CSG.

Speciﬁcation of

user requirements

Output of (1) (3) CTT generation CTT to be validated.

Output of (3) and (1) (4) CTT validation CTT validated.

Output of (4) and (2) (5) CTT Mapping CTT mapped to the conceptual view.

Design of Pro-

posed Solution

Output of (5) (6) Architectural design UI design and navigation ﬂow.

Output of (5) and (6) (7) Structural design Internal structure of each UI.

Output of (5) and (7) (8) Content design UI patterns that compose each UI.

Output of (5) and (8) (9) Reﬁnement Conceptual Design.

Output of (5) and (9) (10) Implementation Implemented solution.

Evaluate Pro-

posed Solution

Output of (10) (11) Validation Validated solution.

plish user goals.

• The GenomIUm Method (I

niguez-Jarrin, 2019):

is a guideline that provides a systematic approach

to designing and implementing UIs for working

with genomics data. It offers a set of design pat-

terns that support the process.

• Technology Acceptance Model (TAM) (Davis,

1989): is an information systems theory used to

model how users accept and use a given technol-

ogy.

The development of Sicila followed the four steps

deﬁned by the ISO 9241-210:2019 (see Table 1 for a

detailed list of activities):

1. Understand and Specify the Context of Use: In

this step, we identify and study the users, their

goals, and the tasks they will perform when us-

ing the system. The CSG is used to understand

and precisely characterize the working domain in

this phase and eases communication between de-

velopers and domain users.

2. Specify User Requirements: The identiﬁcation of

user needs and requirements is carried out in this

phase. We used the CTT formalism to better cap-

ture and describe the speciﬁc tasks identiﬁed in

the previous step that users aim to perform when

using Sibila.

3. Produce Design Solutions: The Design and im-

plementation of Sibila are performed in this

The speciﬁc notation of Concur Task Trees can be

found in https://www.w3.org/2012/02/ctt/

phase. We carry out this process following the

design process described by GenomIUm. We use

the CSG to support the process by mapping data

attributes to the speciﬁc patterns of the UIs.

4. Evaluate Design Against User Requirements: Fi-

nally, we test Sibila against user goals. In this

step, we conducted a TAM-based experiment in

which we measured two metrics: Perceived Ease

of Use (PEOU) and Perceived Usefulness (PU).

3 THE TOOL

In this section, we report how the four phases deﬁned

in the previous section have been carried out to design

and develop Sibila.

3.1 Speciﬁcation of Context of Use

3.1.1 Focus Group

To specify the context of use and discover the user

expectations from the system, we organized several

focus groups over four months with domain experts

to understand in what contexts they would use Sibila.

Sixteen virtual meetings (one per week) and four face-

to-face meetings (one per month) were carried out.

After that, we determined their context of use. This

context can be summarized as a need to ﬁnd rele-

vant variations in the context of precision medicine

(i.e., based on a phenotype or gene of interest, iden-

Human-Centered Design for the Efﬁcient Management of Smart Genomic Information

tify what variations play a key role in both phenotype

expression and gene functionality).

3.1.2 Conceptual View Generation

The next step consisted of creating a conceptual view

(i.e., a subset of elements) of the CSG tailored to the

context of use speciﬁed before by means of the ISGE

method (Garc

ıa S. et al., 2021a)

. Since the CSG pro-

vides a broad, holistic perspective of the genome, it

might contain too much information when applied to

real-world use cases that focus on a speciﬁc domain

dimension. Thus, adopting the CSG in real-world use

cases can be more efﬁcient and straightforward if we

only consider those concepts and relationships that

are relevant to that use case in particular.

We started by presenting the CSG to the domain

users. Then, all together generated a conceptual view

of the CSG with the most relevant concepts tailored to

Sibila’s functionality, which we called CSG-Sibila.

These view offered multiple advantages: i) it im-

proved the communication between the ﬁnal users

and us, ii) it provided a common framework of knowl-

edge for discussion, and iii) it eased data integration.

Fig. 1 shows the resulting conceptual view, which

reduced the number of considered concepts from 60

to 11. Such a reduction in the number of concepts in-

dicated that the working context is signiﬁcantly nar-

rowed. For instance, no information regarding bio-

logical pathways was selected.

In the CSG-Sibila, the central and most important

concept is the VARIATION, identiﬁed by an id and a

name. A variation consists of a change in our DNA

sequence. It is characterized by the reference and al-

ternative alleles (i.e., the value in the DNA sequence

of humans considered “correct” and the altered one

respectively), its type, and the last time it was stud-

ied. Each variation can have a set of HGVS EXPRES-

SIONS (i.e., The HGVS expressions is a way of rep-

resenting DNA variations following a standard pro-

vided by the HUman Genome Organisation (HUGO)

and a set of EXTERNAL ITEMS (i.e., are the appear-

ances of the variation in external data sources). The

EXTERNAL ITEM contains the name, the URL, and the

variation’s speciﬁc identiﬁer in a speciﬁc data source.

DNA VARIATIONS are located in a speciﬁc region

of the DNA sequence. The CSG Sibila represents

the location of VARIATIONS by means of three ap-

praoches. First, the CHROMOSOME where the vari-

ation is located. Second, the set of GENES altered by

The ISGE method allows for generating more narrow

conceptual schemes, called conceptual views, from a more

general one. These conceptual views are tailored to a spe-

ciﬁc use context and ease the adoption of conceptual model-

based techniques.

the variation. Third, the speciﬁc POSITION in the full

genome sequence where the variation is located. In

some cases, the GENES and the speciﬁc POSITION of

a VARIATION can be unknown.

The study of VARIATIONS is relevant because

they are known to be responsible for genetic dis-

eases. The extent to which a VARIATION is responsi-

ble for a genetic disease is known as clinical impact.

VARIATIONS are associated with PHENOTYPES (i.e.,

genetic diseases) through a set of clinical SIGNIFI-

CANCES. A signiﬁcance indicates the pathogenicity

established between a VARIATION and a PHENOTYPE

(e.g., pathogenic, benign, risk factor, etc) and the ev-

idence that supports such assertion (i.e., the method

and the criteria). The ACTIONABILITY concept is

deﬁned as an aggregate calculated from the differ-

ent SIGNIFICANCES of a VARIATION for a PHENO-

TYPE. An ACTIONABILITY is characterized by the

speciﬁc clinical actionability (i.e., disease-causing or

not disease-causing) and the level of evidence used

for such a classiﬁcation (i.e., strong evidence, mod-

erated evidence, limited evidence, or to follow up).

Finally, each VARIATION can contain PHENOTYPE-

related BIBLIOGRAPHY.

3.2 Speciﬁcation of User Requirements

3.2.1 Concur Task Tree Generation

After collecting the initial information, we comple-

mented it by proposing a task-workﬂow to be imple-

mented in Sibila based on the previously identiﬁed

domain users’ needs. This workﬂow was revised with

domain experts, being polished in multiple iterations

until they approved it. As a result, we improved our

understanding of their mental model and generated

Sibila’s task model.

We consolidated the task model of the envisioned

system using the CTT notation (see Fig. 2). The CTT

notation offer several advantages: i) it focuses on the

activities that users aim to perform; ii) it provides a hi-

erarchical structure with a wide range of granularity;

iii) it offers a graphical syntax that is easy to interpret;

iv) it deﬁnes temporal relationships between tasks.

There are three types of tasks in the CTT nota-

tion. The system task deﬁnes tasks that do not require

user interaction (e.g., displaying data) and is depicted

with a monitor. The interaction task deﬁnes tasks that

require user interaction (e.g., ﬁlling out a form) and

is depicted with a hand. Finally, the abstract task de-

ﬁnes higher-level tasks that are decomposed into other

tasks, including abstract, system, or interaction tasks;

this last task is depicted with a cloud.

ENASE 2023 - 18th International Conference on Evaluation of Novel Approaches to Software Engineering

Figure 1: Sibila’s Conceptual view of the CSG.

3.2.2 Concur Task Tree Validation

To validate the CTT (see Fig 2), we mapped the

gathered requirements and needs to the tasks of the

CTT. Then, we discussed such mappings with domain

users, who agreed that the CTT captured their work-

ﬂow appropriately. The validated CTT is composed

of 30 tasks that are distributed in a four-level hierar-

chy.

3.2.3 Concur Task Tree Mapping

After validating the CTT, we mapped its tasks to the

corresponding entities of the CSG-Sibila, which al-

lowed us to better identify the task’s idiosyncrasy and

connect UI requirements to speciﬁc data attributes.

Also, it allowed us to determine the complexity of the

tasks: the more entities, the more complex a task is

expected to be. It also facilitated the selection of the

most appropriate design patterns to display and inter-

act with the data.

3.3 Design of the Proposed Solution

After deﬁning the context of use and identifying

user requirements, the proposed solution can be de-

signed. We focused on transforming the tasks de-

ﬁned in the CTT into a tangible UI design. This

task was supported by the GenomIUm method, a

Pattern-Oriented Design (POD) method developed

by I

niguez-Jarrin that provides solutions for the ge-

nomics domain (I

niguez-Jarrin, 2019).

GenomIUm provides a systematic design process

and a catalog of interconnected patterns to support it.

The catalog divides patterns into four categories, one

per design process step. Some patterns cover gen-

eral design problems (i.e., navigation and component

distribution), while others have domain-speciﬁc de-

sign problems (i.e., genomics data visualization). The

POD approach offers the advantages of using proven

solutions to common user problems in many different

contexts (Javahery et al., 2004).

The GenomIUm systematic design process is

composed of four steps:

1. Architectural Design. The high-level design of

the UI and the navigation ﬂow is performed in

this step. The architectural design is supported by

information architecture patterns, which describe

system-wide solutions to organize the content to

display and establish their interconnections.

2. Structural Design. In this step, the internal struc-

ture of each UI deﬁned in the architectural design

is performed. The structural design is supported

by page patterns, which describe the components

that structure presentation units.

Human-Centered Design for the Efﬁcient Management of Smart Genomic Information

Explore data

(0)

[]

Explore variations

(10)

Explore phenotypes

(11)

Filter

Variations

(20)

Visualize

Variations

(21)

[Edit

visualization*]

(22)

[Select a

variation]

(23)

Visualize

Variation

(24)

[Navigate to variation

detail info*]

(25)

[]>>

|=|

[]>>

[Select

genes]

(30)

[Select

phenotypes]

(31)

|||

[Filter*]

(32)

[Order*]

(33)

|=| |=| |=|

Details

(34)

Affected

Phenotypes

(35)

External

Sources

(36)

[To

phenotypes*]

(37)

[To

Biblio

graphy*]

(39)

[To

External

sources]

(38)

|=| |=|

Visualize

Phenotypes

(26)

Select a

Phenotype

(27)

Visualize

Phenotype

(28)

[Navigate to phenotype

detail info*]

(29)

[]>>

|=||=||=||=|

[]>>

[to

variants*]

(314)

[to

variant]

(315)

[To External

sources]

(316)

Details

(310)

Affected

genes

(311)

Affected

variants

(312)

External

sources

(313)

Figure 2: Generated Concur Task Tree. The ID of each task is depicted between brackets.

Table 2: Deﬁned functional user interfaces.

Functional UI Interaction task ID(s) Application task ID(s)

Home (H) - -

Filter variations (FVS) 20, 30,31 -

Visualize variations (VVS) 22, 23, 32, 33 21

Visualize variation (VV) 25, 37, 38, 39 24, 34, 35, 36

Visualize phenotypes (VPS) 27 26

Visualize phenotype (VP) 29, 314, 315, 316 28, 310, 311, 312, 313

3. Content Design. The selection of the patterns

that conform to the internal structure of eahc UI

is performed in this step, which is supported by

navigational and content patterns that describe the

content components. The objective of these pat-

terns is to allow users to perform a speciﬁc set of

tasks identiﬁed in the CTT.

4. Reﬁnement. The visual details of each pattern

are speciﬁed in this step. This step is required

because, since design patterns provide general UI

design solutions, they have to be adapted to the

particularities of the use case.

3.3.1 Architectural Design

In the architectural design step, the goal is to deﬁne

the UI design of Sibila, which we divided into two

steps, namely, the deﬁnition of the functional UIs (i.e.,

a UI that implements a set of user tasks) and the nav-

igation among them. We deﬁned the functional UIs

by means of a card sorting exercise and a tree test.

In the card sorting exercise, we provided ﬁve domain

users with 27 paper cards (i.e., one per non-abstract

CTT activity) and asked them to group these cards in

a card sorting exercise. Then, we generated a cor-

relation matrix and performed correlation clustering

to calculate the appropriate number of functional UIs

that Sibila should have (see Fig. 3).

As a result, six functional UIs were proposed and

discussed with those domain users that performed the

card sorting exercise. Table 2 shows the resulting

functional UIs, their proposed names, and the tasks

of the CTT they group. This aggregation of tasks into

functional UIs is used to identify the most appropriate

patterns in the following steps.

In the tree test, we provided those domain users

that did not perform the card sorting exercise with

a ﬁve-category tree (one per functional UI obtained

ENASE 2023 - 18th International Conference on Evaluation of Novel Approaches to Software Engineering

Figure 3: Correlation matrix with the identiﬁed tasks.

from the ﬁrst exercise) and a set of ten random CTT

tasks. We asked them to locate each CTT task into

one tree item. After that, we compared the results

with the group obtained in the ﬁrst exercise, getting

100% accuracy. This means that users from the tree

testing exercise grouped the CTT tasks the same way

as users from the card sorting exercise, which indi-

cates that the proposed functional UIs are appropriate

and intuitive.

For deﬁning the navigation among functional UIs,

we connected them based on the hierarchical compo-

sition and temporal operators of the validated CTT.

We also included the data required for navigating

from one functional UI to another. For instance, to

navigate from the “visualize variations” UI to the “vi-

sualize variation” UI, a variation identiﬁer must be se-

lected. Fig. 4 shows the deﬁned navigation between

the functional UIs. We represented it using a directed

graph (DG) since it offers an intuitive representation

of navigation between UIs. Apart from the navigation

shown in Fig. 4, every functional UI can navigate to

the Home UI because they contain a header. The DG

was discussed and validated with domain users in a

workshop.

Figure 4: Directed graph representing Sibila’s navigation.

Additional notes: the “E” node (external) is used to repre-

sent navigation to an external UI; edges are annotated with

the data needed to perform the navigation; annotated edges

with brackets indicate that data is optional; the topbar’s nav-

igation is not represented. H, FVS, VPS, VVS, VP, and VV

correspond to identiﬁers that can be found in Table 2.

3.3.2 Structural Design

In the structural design step, the internal structure of

the identiﬁed functional UIs is deﬁned. We used the

Conceptual Framework Pattern, which promotes de-

signing UIs with the same layout. This layout con-

sisted of three independent sectors: a header contain-

ing the basic navigational functionality, a body con-

taining the components and patterns needed to per-

form the tasks, and a footer (see Fig. 5).

Figure 5: The internal structure of the UIs from Sibila.

3.3.3 Content Design

In the content design step, the goal is to select the

visualization patterns that compose the internal struc-

ture of each functional UI. The Filter variation (FV),

Visualize variations (VVS), Visualize variation (VV),

Visualize phenotypes (VPS), and Visualize phenotype

(VP) UIs have a topbar, a body, and a footer. Regard-

ing the topbar, we selected the Home Link pattern

(Toxboe, 2007). This pattern is a navigation pattern

that allows users to go back to the starting location

(i.e., the Home UI). The content displayed in the func-

tional UIs (i.e., in the body) is organized following

the Card, and the Chunking pattern (Toxboe, 2007),

which eases data visualization. On the one hand, the

card pattern homogenizes the visualization of content

that is heterogeneous; on the other hand, the chunk

pattern groups information into smaller knowledge

units that are processed and remembered more easily.

Unlike the rest of the functional UIs, we used the

reduction pattern (Toxboe, 2007) for designing the

Home (H) UI. This memory and perception pattern

reduces complex behavior to simple tasks, increasing

user performance. We selected this pattern because

domain users stated that they want to start performing

their tasks as fast as possible. They also stated that

in order to increase Sibila’s adoption, the initial per-

ception of complexity should be as little as possible.

Therefore, we simpliﬁed the design of the Home UI

Human-Centered Design for the Efﬁcient Management of Smart Genomic Information

Table 3: Selected patterns in the design of the CD of the SIBILA UI. CTT task 34 corresponds to Details, 35 corresponds to

Affected Phenotypes, and 36 to External Sources.

ID Pattern Applied to CTT Task(s) Entity Attribute

1 Card

Display the name and date of the

variation

34 Variation* name*, date*

2 Card

Display the rest of the variation

information

34, 35, 36 - -

3 Module tabs

Separate the 34, 35, and 36 CTT

tasks into sections that can be

accessed using ﬂat navigation

34, 35, 36 - -

4 Chunking

Group the general information of a

variation

Variation* ref, alt, type*

Chromosome[]* ID*

VariationPosition[] assembly*,start*, end*

Gene[] name*

5 Chunking

Group the HGVS expressions of a

variation

34 HGVS Expression[] expression*

6 Tagging

Label the clinical actionability and

classiﬁcation of a variation for a

phenotype.

35 Actionability[]*

clinicalActionability*,

classiﬁcation*

7 Chunking

Group the information of a

phenotype. Contains 6, 8, and 9

patterns

35 Phenotype[]* name*

8 Chunking

Group the clinical signiﬁcance of a

variation in a phenotype

35 Signiﬁcance[]*

clinicalSigniﬁcance*,

method*, criteria*

9 Chunking

Group the bibliography of a

variation in a phenotype

35 Bibliography[]*

title*, authors*,

date*, url*

10 Chunking

Group the references of a variation

in external datasources

36 ExternalItem[]* ID*, source*, url*

to its minimal expression. As a result, users only have

two options to explore when they access Sibila: DNA

variations or phenotypes.

This step was iterated until domain users con-

ﬁrmed that the functional UIs met their needs and re-

quirements. The UIs have been designed using six-

teen visualization patterns: alternating row color, au-

tocomplete, cards, chart, chunking, genome browser,

genomic ﬁlter, ideogram, module tabs, pagination,

problem summary, rule builder, search, selector of re-

gions, sort by column, and tagging. Each of these

patterns have being instantiated ultiple times accross

the different UIs.

To illustrate how we carried out this step, we show

the selected patterns and the resulting Conceptual De-

sign (CD) of the Visualize Phenotype (VP) functional

UI. Table 3 shows the selected patterns for the VP UI,

their purpose, and the entities of the CSG whose data

is displayed. Fig. 6 shows the resulting CD of the VP

UI.

3.3.4 Reﬁnement

In the reﬁnement step, the patterns selected in the pre-

vious step are adapted to the speciﬁc particularities

of both the data to be displayed and the working use

case. The following design decisions were made in

this step:

• We selected the design guidelines used to imple-

ment the patterns. We followed the design prin-

ciples proposed by Google’s Material

theme to

build a consistent, high-quality UI.

• We mapped each pattern to the attributes of the

CSG that will display using pseudo code. For in-

stance, the pattern with ID 5 displays the expres-

sion attribute of each HGVSEXPRESSION asso-

ciated with the selected variations (see Table 3).

• We deﬁned a set of reusable layouts to determine

how pattern data should be displayed. For in-

stance, the pattern with ID 5 should display each

HGVS expression by using a chip (i.e., a compact

rectangle with rounded borders to represent small

chunks of data) with an outlined style and a back-

ground following the primary color of the palette

deﬁned for Sibila.

• We designed the logo, the color palette, and the

color of the links when user hovers them.

3.3.5 Implementation

Finally, we have implemented Sibila using three-tier

architecture. For the database, we used PostgreSQL,

the backend consists of a node.js server, and the fron-

https://material.io/design

ENASE 2023 - 18th International Conference on Evaluation of Novel Approaches to Software Engineering

Figure 6: Conceptual design of the VP UI.

tend is implemented with JavaScript and the React

framework.

The communication between the database and the

backend uses the knx.js query builder library while

the communication between the backend and the fron-

tend uses the GraphQL query language.

To illustrate, Fig. 7 shows the ﬁnal implementa-

tion of the Visualize Phenotype (VP) UI.

3.4 Evaluate Proposed Solution

3.4.1 Validation

To validate Sibila, we used the Technology Accep-

tance Model (TAM). TAM identiﬁes two perception-

based variables, namely Perceived Ease Of Use

(PEOU) and Perceived Usefulness (PU) (Davis,

1989):

• PEOU: the degree to which users believe that us-

ing Sibila would be free from effort.

Figure 7: The ﬁnal implementation of the VP UI, which is

divided into three tabs. Please note that, since this variation

has no associated bibliography, pattern 9 is not instantiated.

• PU: the degree to which a person believes that us-

ing Sibila would enhance job performance.

The validation was designed by the ten domain

experts involved in the design of Sibila. The do-

main experts prepared two experiments covering all

CTT tasks. The ﬁrst focused on exploring varia-

tions, while the other focused on exploring pheno-

types. Then, the validation was conducted by a group

of ten computer scientists with genomics expertise.

They were divided into two groups of 5 members,

and each group performed one of the prepared experi-

ments. After that, they ﬁlled out a survey with twelve

items for measuring the two TAM variables (see Ta-

ble 4). The survey items were formulated using a 5-

point Likert scale, ranging from 1 (strongly disagree)

to 5 (strongly agree), using the opposing-statement

question format. Besides, the item order was random-

ized, and half of the questions were negated to avoid

Human-Centered Design for the Efﬁcient Management of Smart Genomic Information

0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%

PEOU 1

PEOU 2

PEOU 3

PEOU 4

PEOU 5

PEOU 6

PU1

PU2

PU3

PU4

PU5

PU6

STRONGLY DISAGREE DISAGREE NEUTRAL AGREE STRONGLY AGREE

Figure 8: The validation results.

monotonous responses.

PEOU and PU were calculated as the average of

the responses obtained from the survey. PEOU was

calculated with questions PEOU1 to PEOU6, while

PU was calculated with questions PU1 to PU6. The

PEOU scored 4.62 out of 5, and the PU scored 4.75

out of 5 (see Fig. 8).

4 DISCUSSION

We applied an HCD approach to design a bioinfor-

matics tool called Sibila. The aim of this tool is to

help domain users discover relevant and high-quality

variations affecting speciﬁc phenotypes or genes of

interest. This discovery process is complex and error-

prone because domain users must face a vast amount

of genomics data, some of which fail to achieve the

minimum quality standards. Consequently, the relia-

bility of the data decreases.

Following an HCD approach allowed us to focus

on domain user needs instead of on our assumptions,

which led us to gather more realistic and correct user

requirements. In addition, users’ feedback was col-

lected during the development process rather than just

at the end.

Applying HCD in genomics led to a more com-

plex design and implementation, and it required addi-

tional efforts compared to its application in other do-

mains. Multiple challenges have been faced, includ-

ing the high degree of heterogeneity of genomics data

and the lack of solid and shared domain knowledge

originated from vague and often ambiguous deﬁni-

tions. We could overcome this challenges through the

use of CM techniques. An additional relevant chal-

lenge was data visualization. There are style guides

to visualize data, but there is little literature regarding

the best way to visualize genomics data. For instance,

there are notable differences in how genomics portals

display their data, even though they target the same

audience. However, some common visualization pat-

terns exist, and GenomIUm allowed us to identify and

integrate them into the Sibila UI.

The deﬁnition of domain-speciﬁc concepts and

workﬂows are required to elicit a platform’s require-

ments, but the most elemental domain concepts are

still open to discussion in genomics (Pearson, 2006).

With regard to this, we have found that the use of CM

has been extremely helpful. CM offered us several

beneﬁts and improved the outcomes of an HCD ap-

proach:

• The use of CM forced domain users to make

their implicit knowledge explicit, which generated

fruitful discussions about the dimensions associ-

ated with core deﬁnitions that otherwise would

have remained hidden in domain experts’ minds.

These discussions are relevant because we found

that it is not uncommon for the same concept to

be interpreted differently by domain users. Hav-

ing these discussions solved allowed domain users

to communicate more effectively.

• We detected that the initial gap of knowledge

between domain users and us was reduced after

starting to work with the CSG as it provided a

common framework to discuss in order to get a

shared understanding of the domain under inves-

tigation.

• We were able to perform a more efﬁcient mapping

between the data and the UIs, easing integration.

• We could plan the tool’s implementation better

as CM allowed us to classify each pattern and

UI complexity based on the mapping carried out.

CM has been integral to the development of our

work. Sibila is grounded in the CSG, and most

of the HCD exercises we conducted, such as fo-

cus groups or card sorting, were enriched using

the CSG. Additionally, GenomIUm is a pattern-

oriented solution designed following a CM ap-

proach also.

Another beneﬁt of HCD was increasing teamwork

and collaboration between developers and domain

experts since they were involved in all of the pre-

implementation activities. Thus, we delivered a better

user experience and found that some of our assump-

tions regarding how domain users carry out their tasks

were incorrect. For instance, we initially considered a

different distribution of the data associated with vari-

ations (i.e., including phenotype information on the

main page of the variation). However, users preferred

a different approach (i.e., an independent section with

additional information about the phenotype).

We found that following an HCD approach in-

creased the complexity of the development because

ENASE 2023 - 18th International Conference on Evaluation of Novel Approaches to Software Engineering

Table 4: Validation activity questions.

Item Statement

PEOU1 I believe that it is easy to explore the data and retrieve relevant information

PEOU2 I found it easy to ﬁlter variations.

PEOU3 I believe that the displayed information regarding variations and phenotypes is adequate and

sufﬁcient

PEOU4 I think that the way the data is visualized is useful.

PEOU5 In general, I found Sibila intuitive.

PEOU6 In general, I found Sibila easy to use.

PU1 I believe that the displayed information is clear, concise, and unambiguous.

PU2 I believe that Sibila would reduce the time and effort required to perform my work.

PU3 I believe that Sibila offers an effective solution to visualize variations.

PU4 I believe that Sibila offers an effective solution to explore and exploit variations’ information.

PU5 In general, I think that Sibila would ease my work.

PU6 In general, I think that Sibila is a useful tool.

there was an initial delay that originated from eliciting

user requirements prior to implementation. Besides,

domain users were initially reluctant to discuss using

a CS as a communication tool. This situation delayed

even more user elicitation of requirements because we

had to teach domain users the basics of CM and how

to interpret a CS.

Despite all of this, an HCD approach delivered

UIs perceived by users as easy to use and use-

ful. Also, supporting our process with a conceptual

schema led to more efﬁcient and effective communi-

cation and domain conceptualization.

5 CONCLUSIONS

The correct design of UIs is crucial to managing

data correctly and extracting knowledge efﬁciently.

A well-designed UI can signiﬁcantly impact the per-

formance of these tasks (i.e., data management and

knowledge extraction). In general, genomic tools are

unintuitive and complex. Multiple reasons are behind

this, but the immediate consequence is that learning to

use these tools is difﬁcult and tedious, reducing their

adoption by domain experts.

Our work towards applying HCD techniques em-

phasizes the need for a systematic approach to tackle

this problem by showing how it can help design us-

able UIs. Besides, complementing HCD techniques

with the support of a pattern-based method (i.e.,

GenomIUm) to design the UIs provides signiﬁcant

beneﬁts. On the one hand, HCD allows research users

and better specify their needs and goals. On the other

hand, composing the UIs with widely used and known

patterns eases the adoption of genomics tools and fa-

cilitates domain users’ work.

Future work will provide a broader, empirical val-

idation by comparing Sibila to other known genomic

portals like ClinVar (Stephens et al., 2015) and En-

sembl (Hunt et al., ). This validation will measure

the increase in the user’s performance when using a

pattern-based tool designed using an HCD approach.

ACKNOWLEDGEMENTS

This work was supported by the Valencian Inno-

vation Agency and Innovation through the OG-

MIOS project (INNEST/2021/57), the General-

itat Valenciana through the CoMoDiD project

(CIPROM/2021/023), and the Spanish State Re-

search Agency through the DELFOS (PDC2021-

121243-I00) and SREC (PID2021-123824OB-I00)

projects, MICIN/AEI/10.13039/501 100011033 and

co-ﬁnanced with ERDF and the European Union Next

Generation EU/PRTR. This work also had the support

of the ACIF/2021/117 grant.

REFERENCES

Bolchini, D. et al. (2009). Better bioinformatics through

usability analysis. Bioinformatics, 25(3):406–412.

ISBN: 1367-4803.

Bradwell, H. L. et al. (2019). Companion robots for older

people: importance of user-centred design demon-

strated through observations and focus groups com-

paring preferences of older people and roboticists in

South West England. BMJ Open, 9(9):e032468.

Carpenter, A. E. et al. (2012). A call for bioimaging soft-

ware usability, volume 9. NIH Public Access. ISSN:

15487091 Publication Title: Nature Methods.

Chilana, P. K. et al. (2010). Understanding usability prac-

tices in complex domains. In Conference on Human

Factors in Computing Systems - Proceedings, vol-

Human-Centered Design for the Efﬁcient Management of Smart Genomic Information

ume 4, pages 2337–2346, New York, New York, USA.

ACM Press.

Cutting, E. et al. (2016). User-centered design of multi-gene

sequencing panel reports for clinicians. 63:1–10.

Davis, F. D. (1989). Perceived Usefulness, Perceived Ease

of Use, and User Acceptance of Information Technol-

ogy. MIS Quarterly, 13(3):319.

de Matos, P. et al. (2013). The Enzyme Portal: A case study

in applying user-centred design methods in bioinfor-

matics. BMC Bioinformatics, 14. Publisher: BMC

Bioinformatics.

Dopp, A. R. et al. (2019). Integrating implementation and

user-centred design strategies to enhance the impact

of health services: protocol from a concept mapping

study. Health research policy and systems, 17(1):1–

11. Publisher: Springer.

dos Santos, F. A. et al. (2019). Graphic design and user-

centred design: designing learning tools for primary

school. International Journal of Technology and De-

sign Education, 29(5):999–1009.

Garc

ıa S., A. et al. (2021a). ISGE: A conceptual model-

based method to correctly manage genome data. In

Nurcan, S. and Korthaus, A., editors, Intelligent In-

formation Systems - CAiSE Forum 2021, Melbourne,

VIC, Australia, June 28 - July 2, 2021, Proceedings,

volume 424 of Lecture Notes in Business Information

Processing, pages 47–54. Springer.

Garc

ıa S., A., Palacio, A. L., Rom

an, J. F. R., Casamayor,

J. C., and Pastor, O. (2021b). A conceptual model-

based approach to improve the representation and

management of omics data in precision medicine.

IEEE Access, 9:154071–154085.

Hunt, S. E. et al. Ensembl variation resources. 2018.

niguez-Jarrin, C. (2019). GenomIUm: A Pattern Based

Method for Designing User Interfaces for Genomic

Data Access. PhD Thesis, Universitat Polit

ecnica de

Val

encia.

Jaspers, M. W. (2009). A comparison of usability methods

for testing interactive health technologies: Method-

ological aspects and empirical evidence. International

Journal of Medical Informatics, 78(5):340–353. Pub-

lisher: Elsevier.

Javahery, H. et al. (2004). Beyond Power: Making

Bioinformatics Tools User-centered. Commun. ACM,

47(11):58–63. Place: New York, NY, USA Publisher:

ACM.

Ko, G. H. et al. (2018). Closha: bioinformatics workﬂow

system for the analysis of massive sequencing data.

BMC bioinformatics, 19:43.

Missier, P. et al. (2010). Taverna, reloaded. In Lecture

Notes in Computer Science (including subseries Lec-

ture Notes in Artiﬁcial Intelligence and Lecture Notes

in Bioinformatics), volume 6187 LNCS, pages 471–

481. Springer, Berlin, Heidelberg. ISSN: 03029743.

Patern

o, F. (2003). ConcurTaskTrees: An Engineered No-

tation for Task Models. The Handbook of Task Analy-

sis for Human-Computer Interaction, pages 483–503.

ISBN: 978-0805844337.

Pavelin, K. et al. (2012). Bioinformatics meets user-centred

design: A perspective. PLoS Computational Biology,

8(7):e1002554. Publisher: Public Library of Science.

Pearson, H. (2006). What is a gene? Nature,

441(7092):398–401. Number: 7092 Publisher: Na-

ture Publishing Group.

Recchia, G. et al. (2021). Creating genetic reports

that are understood by nonspecialists: a case study.

22(2):353–361.

Rimmer, J. (2004). Improving software environments

through usability and interaction design, volume 27.

J Audiov Media Med. ISSN: 17453054 Publication

Title: Journal of Visual Communication in Medicine.

Rutherford, P. et al. (2010). Usability of navigation tools

for browsing genetic sequences. In Conferences in Re-

search and Practice in Information Technology Series,

volume 106, pages 33–41. ISSN: 14451336.

Standardization, I. O. f. (2019). ISO 9241-210:2019, Sec-

ond Edition: Ergonomics of human-system interaction

- Part 210: Human-centred design for interactive sys-

tems.

Stephens, Z. D. et al. (2015). Big data: Astronomical or ge-

nomical? PLoS Biology, 13(7):e1002195. Publisher:

Public Library of Science.

Sutcliffe, A. et al. (2010). User engagement by user-centred

design in e-Health. In Philosophical Transactions of

the Royal Society A: Mathematical, Physical and En-

gineering Sciences, volume 368, pages 4209–4224.

Royal Society. ISSN: 1364503X.

Svanæs, D. et al. (2008). The contextual nature of usability

and its relevance to medical informatics. In Studies

in Health Technology and Informatics, volume 136,

pages 541–546. ISSN: 18798365.

Toxboe, A. (2007). User interface design pattern library.

Publication Title: UI Patterns.

Valentin, F. et al. (2010). Fast and efﬁcient searching of

biological data resources-using EB-eye. Brieﬁngs in

Bioinformatics, 11(4):375–384.

ENASE 2023 - 18th International Conference on Evaluation of Novel Approaches to Software Engineering