Expanded abstract

BioData Mining

Mycoplasma Contamination in The 1000 Genomes Project

http://www.cs.ucl.ac.uk/staff/W.Langdon/

Background:

In silco Biology is increasingly important and is often based on
public data.  Biologists increasingly use sophisticated computer tools
to analyse the results not only of their own data but also data given
to them directly by other Biologists.  Publicly funded projects, for
example the 1000 genomes project, have made available vast quantities
of their data.  In fact making data available is often a condition of
either funding or of publication.  
Exponentially growing quantities of such data are now available to
people remote from their original source.  Indeed with time, the data
will increasingly be used by people who have never met the data
provider.  Indeed the original experimenter may have changed jobs or
even retired.
Nonetheless others will still be using the results of their
endeavours.
Indeed the data are freely available and can be used by non-biologists
(eg Computer scientists) who have little experience of common
procedures and problems associated with microbiology wet lab
experiments.


Once data are on the Internet they start to have a life of their own
and may pop up in unexpected locations and be used in unanticipated
ways.  The current Bioinformatics infrastructure already routinely
distributes key datasets (such as the DNA sequences harvested by the
thousand genome project) globally.

While the problem of contamination is well recognised in
microbiology labs the corresponding problem of database corruption has
received less attention.
For example at least one microbe gene has got itself uploaded into the
database holding the reference human genome and is still there.
It has been copied across the planet.  Indeed it has even managed to
get itself incorporated into some Bioinformatics hardware.




Results: 

Bearing the above in mind, a large sample of freely available DNA
sequences from The Thousand Genome Project were down loaded from
Internet.  Each DNA sequence was tested against both the human genome
and against genomes for 30 different species of Mycoplasma bacteria.

Whilst only a fraction of 1000G DNA sequences are from Mycoplasma they
are well spread so that it appears that in excess of seven percent of
1000G samples are contaminated with some DNA from Mycoplasma or a
similar species. 

Many of the DNA sequences are of low quality.  Even so using NCBI
BLAST searches with high quality, high entropy sequences we get the
same result: some 1000G DNA sequences match Mycoplasma but no human
sequences.

Some locations have a higher contamination rate than others, this may
be related to differences in DNA scanner technology between the
different members of the thousand genome project consortium.