Assessing Reproducibility and Value in Genomic Signatures

genomic signatures;prediction;reproducibility;replicability;transparency;rank-based features;covariate adjustment;Biostatistics

Patil, PrasadWolff, Antonio C. ;

Johns Hopkins University

关键词: genomic signatures;    prediction;    reproducibility;    replicability;    transparency;    rank-based features;    covariate adjustment;    Biostatistics;   

Others  :  https://jscholarship.library.jhu.edu/bitstream/handle/1774.2/39737/PATIL-DISSERTATION-2016.pdf?sequence=1&isAllowed=y

瑞士|英语

来源: JOHNS HOPKINS DSpace Repository

【 摘 要 】

Consider a genomic signature to be a set of genes whose measured expression is transformedinto a prediction of an outcome of interest. Such signatures are the bases of a set of FDA-approved medical tests for predicting the risk of distal recurrence and differential survival in breast cancer patients [84, 83, 63]. The goal of these tests is to provide clinicians with an additional piece of prognostic information that may affect their decision making pertaining to the treatment of a breast cancer patient. As it stands, the tests based on these genomic signatures (MammaPrint, Oncotype DX, Prosigna) are not part of the standard of care for a patient, and there are many issues in the translation of these discoveries from bench to bedside that hinder their reliable use [24]. These issues range from insufficiently thorough validation [86], to technical errors or oversights [5,46], to outright retraction of results [75].In addition to the lack of impact in clinical practice, the difficulty of translating these genomic discoveries represents uncertainty about the viability of clinical genomics in general. The vast majority of genetic quantities that are routinely evaluated for a patient were discovered and characterized prior to the era of high-throughput genomics, e.g. [28, 88, 77, 34, 51]. Although there have been one-off successes [89] and a wide range of candidate and pathway discoveries [85], the costs have been great relative to the payout in terms of widespread clinical use [14]. By examining issues at the point of translation, we can begin to provide a clearer picture of what is possible and realistic to accomplish in the clinic with the discoveries that we have from the high-thorughput era.