Higher Chao1 Values (20000-30000)

[KV]

Dear Team, 

I have a fundamental question on Chao1 values. These values were obtained from illumina sequencing  through QIIME analysis and showed a significantly higher than observed OTUS, which is unusual.The sequence numbers were normalized to 44170 sequence reads for all the samples.

I am curious to know whether Chao1 values around 20000  are acceptable. Please check the below table which details alpha diversity measures for the samples. 

 I would like to know whether the data of Chao1 makes sense.  Also, what could be the factor behind this?

 I'll greatly appreciate your kind guidance regarding this issue. 

Thank you in advance.

Kind Regards,

Krishna

Samples	Number of observed OTUs	Richness estimate (Chao1)	Shannon Diversity Index (H)	Simpson Diversity Index (D)	Phylogenetic diversity (PD)	Good’s coverage
1	3449	20017	5.0	0.87	210	0.94
2	3629	17890	4.0	0.65	197	0.93
3	4221	22226	4.8	0.81	230	0.92
4	4083	23825	5.9	0.93	244	0.92
5	4845	26178	6.6	0.93	287	0.91
6	5157	26517	7.1	0.96	309	0.91
7	5566	26663	7.3	0.97	332	0.90
8	5563	29650	6.6	0.94	333	0.90
0	4702	24495	6.1	0.92	275	0.92

[Colin Brislawn]

Hello Krishna,

Thanks for getting in touch with us. 

When I calculate choa1 values, they are almost always higher than observed OTUs because of how the Chao metric works. Here is a pretty good resources that explain how alpha diversity metrics are calculated, and why chao1 is often higher. Basically:

If a sample contains many singletons, it is likely that more undetected OTUs exist, and the Chao 1 index will estimate greater species richness than it would for a sample without rare OTUs.

http://www.coloss.org/beebook/I/gut-symbionts/2/2/4 

Observed: 3449, chao1: 20017

Looks like many OTUs appear in that sample only once, greatly increasing the choa1 metric. 

Let me know if that helps,

Colin

[KV]

Hi Colin, 

Thank you so much for your immediate response. Much helpful.

 I would like to know whether we can represent these data  (chao1) as such or singletons need to be removed for chao1 analysis.

Regards

Krishna

[Colin Brislawn]

Hello Krishna,

I would like to know whether we can represent these data  (chao1) as such or singletons need to be removed for chao1 analysis.

Singletons should not be removed, because chao1 uses those singletons in it's calculations! If you were to filter out singletons, then chao1 == observed species. 

This paper does a good job explaing the chao1 metric: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC93182/ 

"where Sobs is the number of observed species, n1 is the number of singletons (species captured once), and n2 is the number of doubletons (species captured twice)."

I hope this helps,

Colin