Search

Doutoramento em Informática (MAP-i) Submitted by Carla Jesus (cjesus@ua.pt) on 2018-06-15T10:20:46Z No. of bitstreams: 1 Pedro_Sernadela_Tese.pdf: 7916616 bytes, checksum: c72b5a26c87f4257ce56ef067e1e914b (MD5) Made available in DSpace on 2018-06-15T10:20:46Z (GMT). No. of bitstreams: 1 Pedro_Sernadela_Tese.pdf: 7916616 bytes, checksum: c72b5a26c87f4257ce56ef067e1e914b (MD5) Previous issue date: 2018

RD-Connect (rd-connect.eu) is an EU-funded project building an integrated platform to narrow the gaps in rare disease research, where patient populations, clinical expertise and research communities are small in number and highly fragmented. Guided by the needs of rare disease researchers and with neuromuscular and neurodegenerative researchers as its original collaborators, the RD-Connect platform securely integrates multiple types of omics data (genomics, proteomics and transcriptomics) with biosample and clinical information – at the level of an individual patient, a family or a whole cohort, providing not only a centralized data repository but also a sophisticated and user-friendly online analysis system. Whole-genome, exome or gene panel NGS datasets from individuals with rare diseases and family members are deposited at the European Genome-phenome Archive, a longstanding archiving system designed for long-term storage of these large datasets. The raw data is then processed by RD-Connect's standardised analysis and annotation pipeline to make data from different sequencing providers more comparable. The corresponding clinical information from each individual is recorded in a connected PhenoTips instance, a software solution that simplifies the capture of clinical data using the Human Phenotype Ontology, OMIM and Orpha codes. The results are made available to authorised users through the highly configurable online platform (platform.rd-connect.eu), which runs on a Hadoop cluster and uses ElasticSearch – technologies designed to handle big data at high speeds. The user-friendly interface enables filtering and prioritization of variants using the most common quality, genomic location, effect, pathogenicity and population frequency annotations, enabling users from clinical labs without extensive bioinformatics support to do their primary genomic analysis of their own patients online and compare them with other submitted cohorts. Additional tools, such as Exomiser, DiseaseCard, Alamut Functional Annotation (ALFA) and UMD Predictor (umd-predictor.eu) are integrated at several levels. The RD-Connect platform is designed to enable data sharing at various levels depending on user permissions. At the most basic level (“does this specific variant exist in any individual in this cohort?”) it has lit a Beacon within the Global Alliance for Genomics and Health’s Beacon Network (www.beacon-network.org). At the next stage of sharing – finding similarities between patients in different databases with a matching phenotype and a candidate variant in the same gene – it is actively involved in the development of Matchmaker Exchange (www.matchmakerexchange.org), allowing users of different systems to securely exchange information to find confirmatory cases. And finally, since all patients within the system have been consented for data sharing, users of the system, after validation and authorization, are able to access datasets from other centres, providing an instant means of gathering cohorts for cross-validation and further study. Although open to any rare disease, the platform is currently enriched for neuromuscular and neurodegenerative phenotypes and includes almost 1000 genomic datasets from the NeurOmics project (www.rd-neuromics.eu) with several other contributions in the pipeline, including 1000 limb-girdle muscular dystrophy index cases from the Myo-Seq project (www.myo-seq.org) and more. The platform is free of charge to use and is open for contributions of NGS and phenotypic data from research labs worldwide via platform@rd-connect.eu

In recent years, computing allowed an explosion of service provisioning in the cloud. The main paradigm dictates the migration of user information, such as documents, photos and others, from the desktop to the network, allowing anytime, anywhere access through the Internet. This paradigm provided a adequate environment to the emergence of online storage services, such as Amazon S3. This kind of service allows storing digital data in a transparent way, in a pay-as-you-go model. On the other hand, peer-to-peer networks have been responsible for most Internet tra c, mostly derived from the BitTorrent protocol. The inherent support to resource sharing, scalability and fault-tolerance, have been making this approach popular, particularly for le-sharing applications. This thesis presents an implementation of an S3 compatible storage service based on peer-to-peer networks, in particular, through the BitTorrent protocol. Several measurements are discussed, to assess di erences in access time and throughput.