Connecting human voice profiling to genomics: A predictive algorithm for linking speech phenotypes to genetic microdeletion syndromes

Abstract

Changes in vocal acoustic patterns are known to correlate with the occurrence of several diseases and syndromes, many of which do not directly affect the structures or processes that control voice production. In such cases, it is difficult to support the existence of correlated changes in voice. This paper presents a methodology for identifying potential genomic bases for such correlations, by finding links between specific genes involved in the conditions under study, and those involved in voice, speech or language generation. Syndromes associated with chromosomal microdeletions are examined as an illustrative case, with focus on their linkage to the FOXP2 gene which has been strongly implicated in speech and language disorders. A novel path-finding graph algorithm to detect pathway chains that connect the the former to the latter is proposed. Statistical analysis of ensembles of "voice" chains detected by this algorithm indicates that they are predictive of speech phenotypes for the syndromes. Algorithmic findings are validated against clinical findings in the literature pertaining to the actual speech phenotypes that have been found to be associated with these syndromes. This methodology may also potentially be used to predict the existence of voice biomarkers in naive cases where the existence of voice biomarkers has not already been established.