Developing a protocol for bioinformatics analysis: An integrated information behavior and task analysis approach

Abstract

The purpose of this research is to capture, understand, and model the process used by bioinformatics analysts when facing a specific scientific problem. Integrating information behavior with task analysis, we interviewed 20 bioinformatics experts about the process they follow to conduct a typical bioinformatics analysis-a functional analysis of a gene, and then used a task analysis approach to model that process. We found that each expert followed a unique process in using bioinformatics resources, but had significant similarities with their peers. We synthesized these unique processes into a standard research protocol, from which we developed a procedural model that describes the process of conducting a functional analysis of a gene. The model protocol consists of a series of 16 individual steps, each of which specifies detail for the type of analysis, how and why it is conducted, the tools used, the data input and output, and the interpretation of the results. The linking of information behavior and task analysis research is a novel approach, as it provides a rich high-level view of information behavior while providing a detailed analysis at the task level. In this article we concentrate on the latter.

Obstacles and opportunities with using visual and domain-specific languages in scientific programming

“…In addition, scientists might commonly use spreadsheets, which are the most broadly-used end-user programming platform in the world [19], and which play prominent roles in numerous common end-user tasks as well as the work of engineers [15] [18]. Beyond these mass-market DSLs, some studies have mentioned that scientists use highly domain-specific applications (i.e., by direct manipulation in a user interface, rather than by writing code) [1][9], so it is conceivable that they also use certain highly domain-specific programming languages. The goal of our paper is to explore when and how scientists use-or struggle to use-DSLs, thereby revealing opportunities to better support scientific programming.…”

“…For step (1), data were typically stored in simple text files, which sometimes were preprocessed with DSL code. These data were then read in by a traditional language such as Fortran, which was also used to implement step (2) to instantiate and run a model.…”

An agile model-driven method for involving end-users in DSL development

“…As a conclusion of this analysis, we found several unsolved issues despite the full amount of software tools available. This analysis complements the contributions from other authors like [44,9,41,45,46].…”

Modeling and Storing Scientific Protocols

“…Each step of a scientific protocol can be represented by a task [2,3]. To model a protocol, we distinguish its design, that captures its scientific aim, from its implementation, that specifies resources selected to execute the tasks.…”

“…A scientific protocol thus describes how the experiment is conducted and records all necessary information to reproduce the experiment. In bioinformatics, the importance of identifying protocol tasks has been addressed by Stevens et al [2] and Bartlett et al [3], while Tröger [4] has proposed a language for expressing in silico protocols that approximates research method used for in vitro experiments. We propose a high-level abstract model for scientific protocols representing two different layers associated with scientific protocols: design and implementation, and discuss the mapping between them.…”