Performing a research project without the introduction of possible biases has proven to be difficult as artifacts easily is introduced. With respect to this project, there are a few critical appraisals that should be raised owing to their potential to introduce biases.
4.7.1 Technical issues
Although mentioned briefly in the introduction, the potential repercussions of the modifications implemented in order to increase the yield of the nested PCR are worthy of a second reiteration.
First, increased template DNA could have led to an impediment of the polymerase. Second, the increased number of cycles implemented on both water samples and biopsies might have generated non-specific products, which possibly could have led to the designation of these sequences to a novel OTUs. (Kennedy & Oswald 2011) (Nelson et al. 2014) Although not likely to affect the outcome of the most dominating species, these artificial OTUs might have led to other biases such as overestimations of alpha diversity. Third, albeit increased annealing temperature resulted in better primer specificity, there is a risk that the preclusion of eukaryotic sequences in the biopsies was achieved at the expense of the primers being too little sensitive. Thus, poorly characterized sequences often from species typical of environments outside the human body, might have failed to be amplified. Water sample bacteria potentially colonizing the mucosa of our patients could
therefore have gone undetected or below a noticeable level during downstream analysis.
It should be mentioned that since the foundation for the analysis performed in this research project is DNA, there is a chance that several of the sequences designated into OTUs originate from dead,
non-viable bacteria, or remnants of DNA. In fact, it has been proposed that as much as 99% of the bacterial diversity in disinfected waters stem from non-viable or non-culturable bacteria. (Vaz-Moreira et al. 2013) As one of the main aims of this thesis was searching for a potential transmission of bacteria from water to mucosa, there is a chance that a part of the OTUs under investigation to some extent stemmed from non-viable bacteria.
Our current limited knowledge of bacterial species from environments outside the human body might also have posed an impact during the taxonomic designation of the analysed sequences, as relatively little information seem to be available in genome databases as of today. As Greengenes, the database used for this researching purpose, mostly comprise microorganisms associated with human health, the taxonomic designation of less known environmental bacteria might unfortunately have been subject to bias. Problems of taxonomic annotation of drinking water bacteria became particularly evident in a study where 57,6% of the partial 16S rRNA gene sequences could not be classified when using the ribosomal Database Project Classifier. (Revetta et al. 2010) Furthermore, the use of a closed OTU-picking strategy might have led to a failure to identify novel species as sequences not presenting a similarity > 97% to those of the Greengenes database, were excluded from analysis. (Rideout et al. 2014) Consequently, species of novel nature, and possible important drinking water bacteria, might have been precluded from subsequent analysis steps.
4.7.2 Research design
Antecedent to this research, plausible artifacts could have been introduced already during the recruitment of patients. Albeit not presenting any pathological traits of the GI tract, the patients regarded as controls in this research project were initially enrolled to the IBSEN II study due to suspicions of inflammatory conditions. Although disproven to have IBD, there is a possibility that these patients might have other concealed GI illnesses potentially associated with dysbiosis, thus obscuring their value as healthy controls with normobiosis.
A few words with respect to the Jukes-Cantor model should also be mentioned, as this served as the method for identification of matches. As this model does not take into consideration whether the substitution occurs in variable or conserved regions of the sequence, all substitutions are treated equally. Theoretically, if two different alignments present an equal number of substitutions, but where the majority of these substitutions are located in variable and conserved regions respectively, both alignments will obtain the same evolutionary distance by the Jukes-Cantor method. As a result, it is possible that the evolutionary distance of the latter alignment might be slightly underestimated
compared to the distance of the first alignment where an overestimation might occur. Consequently, the identification of matches might have been subject to biases associated with distance
measurements.
OTU matches under investigation in the Fisher exact analyses were submitted to the parameters given in section 2.5.2 in materials and methods. Thus, it should be noted that a change of these parameters such as increases in the requirements to the number of sequences present in a sample, could lead to different results of the statistical analysis.
4.7.3 Mechanisms of contamination
As to plausible mechanisms of contamination seeking to explain the relatively high presence of bacteria associated with human gut in the matches, several possibilities exist. Prior to this research, the distribution systems of several of the patients could have been subject to contamination from sewage. Contamination could also have occurred at the sampling step, as patients took their own samples without supervision of anyone with knowledge of sterile sampling techniques.
Furthermore, as DNA from the water samples already had been extracted and purified before this project, the possibility that contamination of any kind could have occurred during this
pre-processing should not be excluded. Furthermore, pre-processing of materials took place in a lab where microbial research on fecal specimens frequently is performed, and where PCR is extensively used.
Thus, DNA originating from bacteria normally residing within the gut could potentially have been present on benches, equipment, in dust etc and contaminated the samples as most work was performed on a working station in an open environment. Regardless of mechanism of
contamination, as the number of cycles in the first PCR reaction was increased to 30 cycles it is possible that even the smallest traces of contamination could have had an impact on subsequent analysis steps, especially because the amount of DNA in the water samples initially were so low.