Ashleigh Cheyne (United Kingdom)
Imperial College London Life Sciences/MedicinePresenter of 1 Presentation
PERFORMANCE OF HOST BLOOD TRANSCRIPTOMIC SIGNATURES FOR DIAGNOSIS OF PAEDIATRIC TUBERCULOSIS A SOUTH AFRICAN CASE-CONTROL STUDY. (ID 1564)
Abstract
Background
The lack of an accurate diagnostic tests for paediatric tuberculosis (TB) is a major contributing factor to the burden of TB in children. Host blood transcriptomic signatures have shown potential for being used as diagnostic tests. However, most of these signatures were discovered in adult datasets, but their performance has not been assessed in paediatric studies. Here, we perform a comparison of published transcriptomic signatures to assess their potential in distinguishing TB from other diseases (OD) in children.
Methods
117 children with TB and OD were recruited in South Africa between 2009 and 2013 and whole blood RNA-sequencing was performed. After reviewing the literature, we identified 26 transcriptomic signatures that fulfilled our selection criteria on derivation, which were then assessed both using the models described in their original publication and were also refitted to assess their full potential in classifying the patients in our dataset.
Results
Out of the 26 signatures tested using the previously described models, none achieved the optimum WHO Target Product Profile guidelines for sensitivity (>85%) and specificity (>92%) of a novel non-sputum based diagnostic test. However, when we constructed optimised models, 3/26 signatures met the optimal criteria for distinguishing active TB from OD. We observed a relationship between signature size and performance.
Conclusions
Our results highlight that robust and generalisable models for diagnostic transcriptomic signatures are needed to exploit the full potential of gene expression signals measured in blood, accelerating their development into clinically usable diagnostic tests.
Clinical Trial Registration
Not applicable