TY - JOUR
T1 - Robust computational analysis of rRNA hypervariable tag datasets
AU - Sipos, Maksim
AU - Jeraldo, Patricio
AU - Chia, Nicholas
AU - Qu, Ani
AU - Dhillon, A. Singh
AU - Konkel, Michael E.
AU - Nelson, Karen E.
AU - White, Bryan A.
AU - Goldenfeld, Nigel
PY - 2010
Y1 - 2010
N2 - Next-generation DNA sequencing is increasingly being utilized to probe microbial communities, such as gastrointestinal microbiomes, where it is important to be able to quantify measures of abundance and diversity. The fragmented nature of the 16S rRNA datasets obtained, coupled with their unprecedented size, has led to the recognition that the results of such analyses are potentially contaminated by a variety of artifacts, both experimental and computational. Here we quantify how multiple alignment and clustering errors contribute to overestimates of abundance and diversity, reflected by incorrect OUT assignment, corrupted phylogenies, inaccurate species diversity estimators, and rank abundance distribution functions. We show that straightforward procedural optimizations, combining preexisting tools, are effective in handling large (105{106) 16S rRNA datasets, and we describe metrics to measure the effectiveness and quality of the estimators obtained. We introduce two metrics to ascertain the quality of clustering of pyrosequenced rRNA data, and show that complete linkage clustering greatly outperforms other widely used methods.
AB - Next-generation DNA sequencing is increasingly being utilized to probe microbial communities, such as gastrointestinal microbiomes, where it is important to be able to quantify measures of abundance and diversity. The fragmented nature of the 16S rRNA datasets obtained, coupled with their unprecedented size, has led to the recognition that the results of such analyses are potentially contaminated by a variety of artifacts, both experimental and computational. Here we quantify how multiple alignment and clustering errors contribute to overestimates of abundance and diversity, reflected by incorrect OUT assignment, corrupted phylogenies, inaccurate species diversity estimators, and rank abundance distribution functions. We show that straightforward procedural optimizations, combining preexisting tools, are effective in handling large (105{106) 16S rRNA datasets, and we describe metrics to measure the effectiveness and quality of the estimators obtained. We introduce two metrics to ascertain the quality of clustering of pyrosequenced rRNA data, and show that complete linkage clustering greatly outperforms other widely used methods.
UR - http://www.scopus.com/inward/record.url?scp=79251526173&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=79251526173&partnerID=8YFLogxK
U2 - 10.1371/journal.pone.0015220
DO - 10.1371/journal.pone.0015220
M3 - Article
C2 - 21217830
AN - SCOPUS:79251526173
SN - 1932-6203
VL - 5
JO - PloS one
JF - PloS one
IS - 12
M1 - e15220
ER -