Showing posts with label sequencing. Show all posts
Showing posts with label sequencing. Show all posts

Friday, 27 September 2013

Metagenomics: A hypothesis-driven fishing expedition?

Salmonella Typhimurium (red) invading cultured human cells
Rocky Mountain Laboratories, NIAID, NIH
Metagenomics is big news in Bioinformatics. Its rise to fame can be traced back to 2003, when Craig Venter (of private Human Genome Project fame) started sailing around the globe, collecting and sequencing water samples. His pilot project distinguished more than 1800 species, including 148 new types of bacteria [3].

Metagenomics is the study of genetic material recovered directly from environmental samples. There are two common applications: ecological samples (e.g. soil, water, industrial sites) and host-associated microbial communities (e.g. human gut, cow rumen). The metagenomics of microbial communities has been dubbed the microbiome.


Key definitions:
Metagenomics: the study of genetic material recovered directly from environmental samples.
Microbiome: the ecological community of commensal, symbiotic, and pathogenic microorganisms associated with a host organism.


Why do we do metagenomics? According to Scholz [1], we are probably trying to answer one or both of these questions:

Who is there?
What are they doing?

The “who” might be viruses, bacteria, archaea, or small eukaryotes like fungi, protists and plants. If you’re looking at a microbiome, you’re probably going to see plenty of DNA from the host as well. Many studies will focus on one or two of these groups. For example a study of the human gut microbiome, might be most interested in bacterial populations, while a deep-sea vent may be teaming with sulphur-loving archaea.

The goal of the “who” question is ideally to detect which species are present. This detailed level of species classification of often not possible, so most studies will try to detect larger groups, say at the genus, family, or order level. In many cases there will plenty of species that have never been seen before. These will need to be grouped into more manageable categories, often by sequence similarity. Classical taxonomic groupings of sequences are often beyond the computing power of modern metagenomics experiments. More often you would cluster the sequences at a given threshold of sequence similarity and then call the group an operational taxonomic unit (often abbreviated OTU). This gets around the issue of novel sequences popping up, but can get hairy when you are trying to infer an evolutionary relationship between sequences based on similarity.


16S sequencing is a common approach to the question of “who”. There are regions of the 16S ribosomal RNA gene which are remarkably well conserved between species. These sequences can be used to determine phylogenetic relationships between sequences and build up a picture of what types of organisms are present. There are already large databases of 16S sequences which can be used to identify known organisms.

The advantage of the 16S approach is that you can focus in on sequences that are useful for classifying known types of organisms. You also don’t have to worry so much about contamination from the host organism. The disadvantage is that by focusing in like this, you can miss new or unusual sequences. For example primers used to amplify bacterial 16S sequences, will do just that. They will not detect plant, animal or fungal sequences, because they are so diverse from the bacterial ones. However if your goal is to detect known bacteria (or at least bacterial reasonably closely related to known ones), then this is a great strategy. You can also save some money on sequencing if you only need to cover a small region.


An alternative is the shotgun sequencing approach. Here you simply sequence all DNA present in the sample. Here your sequencing is no longer biased by the primers you choose. You simply see everything. Of course you are still more likely to sequence more abundant species. You will probably have to do quite a lot more sequencing to detect the same number of species compared with 16S because your sequence could come from any (potentially uninformative) part of the genome.
Shotgun Sequencing - Wooley JC, Godzik A, Friedberg I (2010) A Primer on Metagenomics. PLoS Comput Biol 6(2): e1000667. doi:10.1371/journal.pcbi.1000667
With shotgun sequencing the question of who is there can become a little trickier. One approach is to look for representative genes. Find the genes that make one species different from another and you have a way to identify them. In general you will have more trouble figuring out who is there, but suddenly you have a rich source of information about what they are doing.

So “what are they doing?” We have answered our question about which species or operational taxonomic units are there, now we want to know what they are up to. We might be interested in particular gene pathways, for example if we pick up an ocean sample near an oil spill we may be looking for genes involved in metabolising hydrocarbons. We may be interested in pathways that span species, where groups of organisms are working in symbiosis. One way to find such genes is to map our reads against databases of enzymes such as KEGG http://www.genome.jp/kegg/. These are already well annotated in terms of function.


So what is the best way answer the two questions: Who is there? What are they doing? Should you do 16S sequencing or shotgun? It really depends on which question is most important to you and much much funding you have! But here are some of the main points to consider:

16S
  • Limited to identifying taxa 
  • Requires longer reads (e.g. 454) 
  • Can miss organisms when primers don’t bind 
  • Biased towards known species 
  • Don’t amplify host DNA when studying host-associated environments

Shotgun
  • Can do gene/pathway level analysis 
  • Can use shorter reads (e.g. Illumina) 
  • Preferably use paired end for assembly 
  • Requires higher depth to detect the same number of species compared with 16S Host DNA contamination 
  • More computationally intensive

References

  1. Scholz, Matthew B., Chien-Chi Lo, and Patrick SG Chain. "Next generation sequencing and bioinformatic bottlenecks: the current state of metagenomic data analysis." Current opinion in biotechnology 23.1 (2012): 9-15.
    http://www.sciencedirect.com/science/article/pii/S0958166911007245
  2. The human microbiome project (2012). Structure, function and diversity of the healthy human microbiome. Nature, 486, 207-214.
    http://eprints.ugd.edu.mk/2482/
  3. http://www.sciencemag.org/content/304/5667/66
  4. More on ‘omics: http://www.the-scientist.com/?articles.view/articleNo/13313/title/-Ome-Sweet--Omics---A-Genealogical-Treasury-of-Words/