Events in early life contribute to subsequent risk of asthma; however, the causes and trajectories of childhood wheeze are heterogeneous and do not always result in asthma. Similarly, not all atopic individuals develop wheeze, and vice versa. The reasons for these differences are unclear. Using unsupervised model-based cluster analysis, we identified latent clusters within a prospective birth cohort with deep immunological and respiratory phenotyping. We characterised each cluster in terms of immunological profile and disease risk, and replicated our results in external cohorts from the UK and USA. We discovered three distinct trajectories, one of which is a high-risk 'atopic' cluster with increased propensity for allergic diseases throughout childhood. Atopy contributes varyingly to later wheeze depending on cluster membership. Our findings demonstrate the utility of unsupervised analysis in elucidating heterogeneity in asthma pathogenesis and provide a foundation for improving management and prevention of childhood asthma.
This study utilises extensive data from human subjects, specifically paediatric cohorts, for which eLife's policies recognise that there can be strong reasons to restrict access. For each of the cohorts involved in our study (CAS, COAST, MAAS), parents were consented on the use of biomedical data for allergy and asthma research, but not for the open sharing of their or their children's data. Studies were run in the late 1990s and early 2000s and we do not have ethics permission to attempt to recontact families to seek consent. Importantly, we note that key data features could risk re-identification of subjects (e.g. demographic data from small communities).However, we have provided public data at the summary level which can be used for subsequent studies, such as replication and meta-analysis. This is standard practice in sensitive data settings, such as genome-wide association studies. These data have been uploaded as Excel spreadsheets to FigShare for ease of data extraction:Supplementary Table 4 https://figshare.com/articles/Supplementary_File_1_1/6934052Supplementary Table 7 https://figshare.com/articles/Supplementary_File_1_2/6934055
- Michael Inouye
- Howard HF Tang
The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.
Human subjects: Ethics approval and consent requirements for each cohort were met as follows: The CAS study was approved by the ethics committees of the King Edward Memorial and Princess Margaret Hospitals in Western Australia; fully informed parental consent was obtained for all subjects. The COAST study was approved by the Human Subjects Committee of the University of Wisconsin. The MAAS study was approved by a Manchester Local Research Ethics Committee (ERP/94/032; SOU/00/258; 03/SM/400; Study registration ISRCTN72673620); fully informed parental consent was obtained for all subjects across all cohorts.
- M Dawn Teare, University of Sheffield, United Kingdom
© 2018, Tang et al.
This article is distributed under the terms of the Creative Commons Attribution License permitting unrestricted use and redistribution provided that the original author and source are credited.
Studies of protein fitness landscapes reveal biophysical constraints guiding protein evolution and empower prediction of functional proteins. However, generalisation of these findings is limited due to scarceness of systematic data on fitness landscapes of proteins with a defined evolutionary relationship. We characterized the fitness peaks of four orthologous fluorescent proteins with a broad range of sequence divergence. While two of the four studied fitness peaks were sharp, the other two were considerably flatter, being almost entirely free of epistatic interactions. Mutationally robust proteins, characterized by a flat fitness peak, were not optimal templates for machine-learning-driven protein design – instead, predictions were more accurate for fragile proteins with epistatic landscapes. Our work paves insights for practical application of fitness landscape heterogeneity in protein engineering.
Using a neural network to predict how green fluorescent proteins respond to genetic mutations illuminates properties that could help design new proteins.