Hand bacteria as an identifier: a biometric evaluation

• Published in: Network modeling and analysis in health informatics and bioinformatics (Wien) 2015-12, Vol.4 (1), p.22, Article 22
• Main Authors: Holbert, Amanda B., Whitelam, Holly P., Sooter, Letha J., Hornak, Larry A., Dawson, Jeremy M.
• Format: Article
• Language: English
• Subjects: Algorithms; Applications of Graph Theory and Complex Networks; Bacteria; Bioinformatics; Biomarkers; Biometrics; Clustering; Computational Biology/Bioinformatics; Computer Science; Data collection; Demographics; Deoxyribonucleic acid; DNA; Ethnicity; Forensic sciences; Gender; Health Informatics; Hispanic Americans; Identification; Keyboards; Microorganisms; Minority & ethnic groups; Odors; Original Article; Population; rRNA 16S; Skin; Support vector machines; VOCs; Volatile organic compounds
• ISSN: 2192-6662; 2192-6670
• DOI: 10.1007/s13721-015-0095-0

Molecular and soft bio-molecular biometrics are an advancing field that involves the analysis of a person’s unique biological markers at a molecular level to ascertain identity. Bacteria communities found on the skin of the human hand have shown to be highly diverse and to have a low percentage of similarity between individuals. The goal of this research effort is to see if a person’s demographics, primarily ethnicity, share a relationship with the bacteria communities that exist on their hand. A sample collection was carried out in which the left and right inner palms of 250 individuals were swabbed to obtain a total of 500 bacteria samples. Of these, 104 samples from 52 individuals (left and right hands) covering a range of age, gender, and ethnicity of the participants were sequenced using 150 paired-end multiplex reads on an Illumina MiSeq. The reads contained the third hypervariable region DNA of the microbial 16S rRNA gene commonly used for microbial identification. Sequences were analyzed using a combination of commercial and custom bioinformatics tools. Results indicated that women who participated in the sample collection had a 15.7 % higher diversity of bacteria at the genus level than men. Using a support vector machine with a 60 % train and 40 % test approach, ethnicities of individuals who provided samples could be classified with a range of 72–94 % accuracy depending on the method used. Principal coordinate plots generated using the unique fraction (UniFrac) algorithm devised by Lozupone et al. at University of Colorado at Boulder showed that similar clustering appeared with people of Turkish, Asian Indian, and Middle Eastern descent and less clustering with people of Caucasian and African American descent. Although focused on a small subset of the human population with no temporal variance in bacterial diversity explored, these results provide a basis for performing identification based on human bacteria that can be expanded upon using time varying sampling and other regions of the 16S rRNA gene.