Predicting Hypertension Subtypes with Machine Learning Using Targeted Metabolites and Their Ratios.

Hypertension is a major global health problem with high prevalence and complex associated health risks. Primary hypertension (PHT) is most common and the reasons behind primary hypertension are largely unknown. Endocrine hypertension (EHT) is another complex form of hypertension with an estimated prevalence varying from 3 to 20% depending on the population studied. It occurs due to underlying conditions associated with hormonal excess mainly related to adrenal tumours and sub-categorised: primary aldosteronism (PA), Cushing's syndrome (CS), pheochromocytoma or functional paraganglioma (PPGL). Endocrine hypertension is often misdiagnosed as primary hypertension, causing delays in treatment for the underlying condition, reduced quality of life, and costly antihypertensive treatment that is often ineffective. This study systematically used targeted metabolomics and high-throughput machine learning methods to predict the key biomarkers in classifying and distinguishing the various subtypes of endocrine and primary hypertension. The trained models successfully classified CS from PHT and EHT from PHT with 92% specificity on the test set. The most prominent targeted metabolites and metabolite ratios for hypertension identification for different disease comparisons were C18:1, C18:2, and Orn/Arg. Sex was identified as an important feature in CS vs. PHT classification.

Collaborative improvement in Scottish GP clusters after the Quality and Outcomes Framework: a qualitative study.

BACKGROUND: Scotland abolished the Quality and Outcomes Framework (QOF) in April 2016, before implementing a new Scottish GP contract in April 2018. Since 2016, groups of practices (GP clusters) have been incentivised to meet regularly to plan and organise quality improvement (QI) as part of this new direction in primary care policy. AIM: To understand the organisation and perceived impact of GP clusters, including how they use quantitative data for improvement. DESIGN AND SETTING: Thematic analysis of semi-structured interviews with key stakeholders (n = 17) and observations of GP cluster meetings (n = 6) in two clusters. METHOD: This analytical strategy was combined with a purposive (variation) sampling approach to the sources of data, to try to identify commonalities across diverse stakeholder experiences of working in or on the idea of GP clusters. Variation was sought particularly in terms of stakeholders' level of involvement in improvement initiatives, and in their disciplinary affiliations. RESULTS: There was uncertainty as to whether GP clusters should focus on activities generated internally or externally by the wider healthcare system (for example, from Scottish Health Boards), although the two observed clusters generally generated their own ideas and issues. Clusters operated with variable administrative/managerial and data support, and variable baseline leadership experience and QI skills. Qualitative approaches formed the focus of collaborative learning in cluster meetings, through sharing and discussion of member practices' own understandings and experiences. Less evidence was observed of data analytics being championed in these meetings, partly because of barriers to accessing the analytics data and existing data quality. CONCLUSION: Cluster development would benefit from more consistent training and support for cluster leads in small-group facilitation, leadership, and QI expertise, and data analytics access and capacity. While GP clusters are up and running, their impact is likely to be limited without further investment in developing capacity in these areas.

SNAPPI-DB: a database and API of Structures, iNterfaces and Alignments for Protein-Protein Interactions.

SNAPPI-DB, a high performance database of Structures, iNterfaces and Alignments of Protein-Protein Interactions, and its associated Java Application Programming Interface (API) is described. SNAPPI-DB contains structural data, down to the level of atom co-ordinates, for each structure in the Protein Data Bank (PDB) together with associated data including SCOP, CATH, Pfam, SWISSPROT, InterPro, GO terms, Protein Quaternary Structures (PQS) and secondary structure information. Domain-domain interactions are stored for multiple domain definitions and are classified by their Superfamily/Family pair and interaction interface. Each set of classified domain-domain interactions has an associated multiple structure alignment for each partner. The API facilitates data access via PDB entries, domains and domain-domain interactions. Rapid development, fast database access and the ability to perform advanced queries without the requirement for complex SQL statements are provided via an object oriented database and the Java Data Objects (JDO) API. SNAPPI-DB contains many features which are not available in other databases of structural protein-protein interactions. It has been applied in three studies on the properties of protein-protein interactions and is currently being employed to train a protein-protein interaction predictor and a functional residue predictor. The database, API and manual are available for download at: http://www.compbio.dundee.ac.uk/SNAPPI/downloads.jsp.

The contrasting properties of conservation and correlated phylogeny in protein functional residue prediction.

BACKGROUND: Amino acids responsible for structure, core function or specificity may be inferred from multiple protein sequence alignments where a limited set of residue types are tolerated. The rise in available protein sequences continues to increase the power of techniques based on this principle. RESULTS: A new algorithm, SMERFS, for predicting protein functional sites from multiple sequences alignments was compared to 14 conservation measures and to the MINER algorithm. Validation was performed on an automatically generated dataset of 1457 families derived from the protein interactions database SNAPPI-DB, and a smaller manually curated set of 148 families. The best performing measure overall was Williamson property entropy, with ROC0.1 scores of 0.0087 and 0.0114 for domain and small molecule contact prediction, respectively. The Lancet method performed worse than random on protein-protein interaction site prediction (ROC0.1 score of 0.0008). The SMERFS algorithm gave similar accuracy to the phylogenetic tree-based MINER algorithm but was superior to Williamson in prediction of non-catalytic transient complex interfaces. SMERFS predicts sites that are significantly more solvent accessible compared to Williamson. CONCLUSION: Williamson property entropy is the the best performing of 14 conservation measures examined. The difference in performance of SMERFS relative to Williamson in manually defined complexes was dependent on complex type. The best choice of analysis method is therefore dependent on the system of interest. Additional computation employed by Miner in calculation of phylogenetic trees did not produce improved results over SMERFS. SMERFS performance was improved by use of windows over alignment columns, illustrating the necessity of considering the local environment of positions when assessing their functional significance.

A comparison of SCOP and CATH with respect to domain-domain interactions.

The analysis and prediction of protein-protein interaction sites from structural data are restricted by the limited availability of structural complexes that represent the complete protein-protein interaction space. The domain classification schemes CATH and SCOP are normally used independently in the analysis and prediction of protein domain-domain interactions. In this article, the effect of different domain classification schemes on the number and type of domain-domain interactions observed in structural data is systematically evaluated for the SCOP and CATH hierarchies. Although there is a large overlap in domain assignments between SCOP and CATH, 23.6% of CATH interfaces had no SCOP equivalent and 37.3% of SCOP interfaces had no CATH equivalent in a nonredundant set. Therefore, combining both classifications gives an increase of between 23.6 and 37.3% in domain-domain interfaces. It is suggested that if possible, both domain classification schemes should be used together, but if only one is selected, SCOP provides better coverage than CATH. Employing both SCOP and CATH reduces the false negative rate of predictive methods, which employ homology matching to structural data to predict protein-protein interaction by an estimated 6.5%.

Biological units and their effect upon the properties and prediction of protein-protein interactions.

Structural data as collated in the Protein Data Bank (PDB) have been widely applied in the study and prediction of protein-protein interactions. However, since the basic PDB Entries contain only the contents of the asymmetric unit rather than the biological unit, some key interactions may be missed by analysing only the PDB Entry. A total of 69,054 SCOP (Structural Classification of Proteins) domains were examined systematically to identify the number of additional novel interacting domain pairs and interfaces found by considering the biological unit as stored in the PQS (Protein Quaternary Structure) database. The PQS data adds 25,965 interacting domain pairs to those seen in the PDB Entries to give a total of 61,783 redundant interacting domain pairs. Redundancy filtering at the level of the SCOP family shows PQS to increase the number of novel interacting domain-family pairs by 302 (13.3%) from 2277, but only 16/302 (1.4%) of the interacting domain pairs have the two domains in different SCOP families. This suggests the biological units add little to the elucidation of novel biological interaction networks. However, when the orientation of the domain pairs is considered, the PQS data increases the number of novel domain-domain interfaces observed by 1455 (34.5%) to give 5677 non-redundant domain-domain interfaces. In all, 162/1455 novel domain-domain interfaces are between domains from different families, an increase of 8.9% over the PDB Entries. Overall, the PQS biological units provide a rich source of novel domain-domain interfaces that are not seen in the studied PDB Entries, and so PQS domain-domain interaction data should be exploited wherever possible in the analysis and prediction of protein-protein interactions.