Under-Powered, Quick and Dirty Metagenomic Investigation of Plasma - ME/CFS -ADCLS - etc.

Topics with information and discussion about published studies related to Lyme disease and other tick-borne diseases.
Post Reply
Posts: 294
Joined: Sun 7 Apr 2013 15:36

Under-Powered, Quick and Dirty Metagenomic Investigation of Plasma - ME/CFS -ADCLS - etc.

Post by dlf » Mon 7 Nov 2016 16:38

Metagenomic Investigation of Plasma in Individuals with ME/CFS Highlights the Importance of Technical Controls to Elucidate Contamination and Batch Effects

Ruth R. Miller, Miguel Uyaguari-Diaz, Mark N. McCabe, Vincent Montoya, Jennifer L. Gardy, Shoshana Parker, Theodore Steiner, William Hsiao, Matthew J. Nesbitt, Patrick Tang, David M. Patrick , for the CCD Study Group
Published: November 2, 2016http://dx.doi.org/10.1371/journal.pone.0165691

http://journals.plos.org/plosone/articl ... ne.0165691

Myalgic Encephalomyelitis/Chronic Fatigue Syndrome (ME/CFS) is a debilitating disease causing indefinite fatigue. ME/CFS has long been hypothesised to have an infectious cause; however, no specific infectious agent has been identified. We used metagenomics to analyse the RNA from plasma samples from 25 individuals with ME/CFS and compare their microbial content to technical controls as well as three control groups: individuals with alternatively diagnosed chronic Lyme syndrome (N = 13), systemic lupus erythematosus (N = 11), and healthy controls (N = 25). We found that the majority of sequencing reads were removed during host subtraction, thus there was very low microbial RNA content in the plasma. The effects of sample batching and contamination during sample processing proved to outweigh the effects of study group on microbial RNA content, as the few differences in bacterial or viral RNA abundance we did observe between study groups were most likely caused by contamination and batch effects. Our results highlight the importance of including negative controls in all metagenomic analyses, since there was considerable overlap between bacterial content identified in study samples and control samples. For example, Proteobacteria, Firmicutes, Actinobacteria, and Bacteriodes were found in both study samples and plasma-free negative controls. Many of the taxonomic groups we saw in our plasma-free negative control samples have previously been associated with diseases, including ME/CFS, demonstrating how incorrect conclusions may arise if controls are not used and batch effects not accounted for.

Some snips about the Borrelia findings ( in plasma) and the "Alternately Diagnosed Chronic Lyme" patients that I think need more and better characterized analysis...........
Of note, Kraken analysis identified RNA from Borrelia, the bacterial genus that causes Lyme disease, in 16 samples, of which three were from participants with ADCLS. Further investigation of the reads matching Borrelia RNA revealed that only five mapped to Borrelia in both reads from the pair, of which only one pair was from a participant with ADCLS. Furthermore, using BLASTn, with the same parameters used for viral analysis against the NCBI nr database, revealed only one single read out of all 47 identified by Kraken matched to Borrelia, and this was from a healthy participant. Therefore, we were unable to find evidence for the presence of Borrelia in plasma from participants with ADCLS.
We chose to analyse the RNA content of our samples, meaning that our results represent a composite picture combining the actual cellular abundance with metabolic and replicative activity of the microbes present. This therefore may not entirely reflect the bacterial abundance of each sample. Despite this limitation, batch-to-batch effects and sample contamination were both clearly identified, highlighting their strong impact. It should be noted that it is possible that differences between study groups would have been identified if we had performed DNA analysis, which is as a more precise measure of cellular abundance, however, the large batch and contamination effects, as well as the low levels of microbial RNA present, make this highly unlikely.
It is also possible that some taxonomies identified could be false positives occurring due to incorrect alignments made by Kraken or BLASTn. Using a bacterial mock community, we calculated that our bioinformatics pipeline has sensitivity and specificity of 98%, suggesting that 2% of reads may be false positive or false negative alignments. This is particularly likely for bacterial genera seen at low levels, for example Borrelia, which was proven to consist of largely incorrect alignments when validated using BLASTn. Similarly, for viral alignments, Orthobunyavirus RNA was found in 84/102 (82%) of samples (data not shown), however, Orthobunyavirus is often identified due to non-specific alignments of bacterial sequences to viral databases, thus may not truly be present in our samples. It is likely that aligning all reads to the complete NCBI database using BLASTn would reduce false positive taxonomic assignments, however, such an analysis would be highly time consuming, hence alternative faster alignment methods were chosen.
The lack of a positive association between ME/CFS and ADCLS and plasma microbial content in this study does not mean that these syndromes do not have an infectious cause, and future experiments should examine the microbiome of other body tissues, including the cellular blood microbiome, which may harbour higher concentrations of microbes. The gut microbiome is of particular interest, as it has been associated with other chronic diseases, including for example, inflammatory bowel disease [33, 34], asthma [35], and multiple sclerosis [36]. In the ME/CFS field, 16S rRNA gene sequencing of stool samples from ME/CFS patients and healthy controls demonstrated a significant increase in Lactonifactor and decrease in Holdemania in ME/CFS patients compared to controls [37]. A second study also using 16S rRNA gene sequencing of stool samples found a significant decrease in Actinobacteria and one other phylum in ME/CFS patients compared to controls, but no significant differences in bacterial phyla from blood samples. However, both ME/CFS studies were small, with sample sizes of 43 and 10 ME/CFS patients respectively. Furthermore, neither study used a negative control, so they were unable to rule out contamination as a source of their differences. Thus, confirmation of these results requires replication in other populations and laboratories, with careful use of control samples.

Our study was unable to find a positive association between the plasma RNA content of individuals with ME/CFS or ACDLS and healthy controls or controls with SLE. Plasma represents a difficult medium for metagenomic analysis, since very low levels of microbes mean it is very difficult to distinguish the microbiome from background contamination. This is not to say that associations with ME/CFS or other syndromes might not be identified in the microbiome of other body sites, or perhaps in the investigation of host factors such as host gene expression or immunological profiles. However, future exploration must be performed rigorously with large sample sizes, clear definitions and careful use of positive and negative controls.
It would be helpful if someone technically savvy could explain the deficiencies of the authors' hypothesis, methodology and inferences and maybe why this study was not a complete waste of time and resources.

Post Reply