Bioinformatic workflow fragment discovery leveraging the social-aware knowledge graph

Abstract

Constructing a novel bioinformatic workflow by reusing and repurposing fragments crossing workflows is regarded as an error-avoiding and effort-saving strategy. Traditional techniques have been proposed to discover scientific workflow fragments leveraging their profiles and historical usages of their activities (or services). However, social relations of workflows, including relations between services and their developers have not been explored extensively. In fact, current techniques describe invoking relations between services, mostly, and they can hardly reveal implicit relations between services. To address this challenge, we propose a social-aware scientific workflow knowledge graph (S2KG) to capture common types of entities and various types of relations by analyzing relevant information about bioinformatic workflows and their developers recorded in repositories. Using attributes of entities such as credit and creation time, the union impact of several positive and negative links in S2KG is identified, to evaluate the feasibility of workflow fragment construction. To facilitate the discovery of single services, a service invoking network is extracted form S2KG, and service communities are constructed accordingly. A bioinformatic workflow fragment discovery mechanism based on Yen’s method is developed to discover appropriate fragments with respect to certain user’s requirements. Extensive experiments are conducted, where bioinformatic workflows publicly accessible at the myExperiment repository are adopted. Evaluation results show that our technique performs better than the state-of-the-art techniques in terms of the precision, recall, and F1.

MANSOR: A module alignment method based on neighbor information for scientific workflow

“…Specifically, WordNet is used to calculate the similarity of the modules' names, while the xsimilarity algorithm 39 is utilized to compute the similarity of their descriptions. The semantic similarity between modules is then determined by taking the weighted average of the two aforementioned similarities. Domain information‐based approach (DIA) 26 : In this method, domain‐related information including descriptions and tags of workflows corresponding to modules are utilized besides modules' names and descriptions. Leveraging those features, the similarity of modules is computed by the cosine similarity of vectors generated by the Latent Dirichlet Allocation model 40 …”

“…Domain information‐based approach (DIA) 26 : In this method, domain‐related information including descriptions and tags of workflows corresponding to modules are utilized besides modules' names and descriptions. Leveraging those features, the similarity of modules is computed by the cosine similarity of vectors generated by the Latent Dirichlet Allocation model 40 …”

“…Interoperability of services can be mined from related ontology descriptions, 22,23 but this approach relies on the existence of well‐curated ontologies and authors using these ontologies to tag their modules 24 . To address these limitations, the authors of References 25 and 26 propose a Service Social Knowledge Graph to integrate and manage social information between services and developers. Additionally, a connection‐aware reputation score is computed to assist in selecting candidate services.…”