PaBio 16s full length results

[Rainjie]

Hello everyone, I used DADA2 to denoise the PaBio 16s full length results, and it shows that only about 52% of the reads remain, meaning approximately half of the reads were filtered out. My question is there any way to improve non-chimeric reads?

Below are my run parameters(almost default parameters due to good sequence quality ):
--p-min-len 1000 \
p-max-len 1600 \
-p-max-ee 2 \
-p-chimera-method consensus \
--p-trunc-len 0 \
-p-front AGRGTTYGATYMTGGCTCAG --p-adapter RGYTACCTTGTTACGACTT \
-p-pooling-method pseudo

sample-id	input	primer-removed	percentage of input primer-removed	filtered	percentage of input passed filter	denoised	non-chimeric	percentage of input non-chimeric
#q2:types	numeric	numeric	numeric	numeric	numeric	numeric	numeric	numeric
TEST-2-1	110107	96818	87.93	68197	61.94	66051	64843	58.89
TEST-2-10	91515	78719	86.02	55035	60.14	52532	52285	57.13
TEST-2-2	106974	91424	85.46	63453	59.32	60568	60245	56.32
TEST-2-3	79565	67482	84.81	46097	57.94	41891	41196	51.78
TEST-2-4	130730	118701	90.8	82551	63.15	80004	79610	60.9
TEST-2-5	100513	87518	87.07	60225	59.92	56492	54668	54.39
TEST-2-6	65963	55426	84.03	36095	54.72	30673	29070	44.07
TEST-2-7	62975	53712	85.29	35325	56.09	29044	27983	44.44
TEST-2-8	113017	99407	87.96	68709	60.8	66229	65375	57.85
TEST-2-9	62697	52938	84.43	34858	55.6	28950	28095	44.81
Control1	122802	105774	86.13	73771	60.07	70371	69612	56.69
Control10	65815	56335	85.6	36811	55.93	32124	31870	48.42
Control2	87008	74489	85.61	50744	58.32	46378	46034	52.91
Control3	88636	75912	85.64	49556	55.91	44859	44449	50.15
Control4	90676	78532	86.61	52777	58.2	48690	48496	53.48
Control5	98532	84166	85.42	56824	57.67	52006	51682	52.45
Control6	132235	113503	85.83	80783	61.09	79550	79170	59.87
Control7	57607	48528	84.24	30199	52.42	25273	25053	43.49
Control8	90543	76989	85.03	51824	57.24	47621	47056	51.97
Control9	69478	59249	85.28	38333	55.17	34027	33461	48.16
Model1	102228	87159	85.26	60575	59.25	58187	57810	56.55
Model10	79340	66546	83.87	45283	57.07	41115	40382	50.9
Model2	124623	106422	85.4	74589	59.85	72593	72375	58.08
Model3	58893	50807	86.27	33940	57.63	26314	25334	43.02
Model4	108932	92409	84.83	64621	59.32	61939	61535	56.49
Model5	77426	66561	85.97	46222	59.7	42811	42212	54.52
Model6	128813	114801	89.12	79910	62.04	77416	75340	58.49
Model7	130817	111453	85.2	78068	59.68	74622	73891	56.48
Model8	55531	46416	83.59	31201	56.19	27240	26667	48.02
Model9	69684	58807	84.39	38825	55.72	33134	32213	46.23

Here is 16s full length sequence quailty.
forward-seven-number-summaries.tsv (162.4 KB)

[SoilRotifer (Mike Robeson)]

Hi @Rainjie,

Altering the value for --p-min-fold-parent-over-abundance might help. See here.

[Rainjie]

Hi, thanks for your reply. However, adding the parameter --p-min-fold-parent-over-abundance 8/16 does not improve the results. Still nearly half of the reads did not pass the chimeric filter.
Any suggestion?

[SoilRotifer (Mike Robeson)]

Hi @Rainjie,

Sadly, I've never worked with PacBio data before, so that was my only quick suggestion. Not sure what the quality plots look like, but would truncation help at all? Perhaps someone in the forum with more experience working with PacBio data can help here?

-Mike

[Rainjie]

Hi Mike,

Thank you for your quick reply and for the suggestion about truncation. I really appreciate you taking the time to think about it. I’ll keep experimenting and also see if anyone else with PacBio experience can offer additional advice.

Thanks again, and have a nice day!

[Oddant1 (Anthony Simard)]

Hello @Rainjie,

Do you still require additional assistance with this?

[Rainjie]

Yeah, Any other suggestions?
--p-min-fold-parent-over-abundance 8/16 does not improve the results.

[Oddant1 (Anthony Simard)]

Can you please run qiime demux summarize on your data and post the resulting visualization here? That will help us figure out where your parameters should be set.

[Rainjie]

I’m using dada2 denoise-ccs. Given the expected amplicon length of 16s rRNA sequence, I’m currently using:
--p-min-len 1000 \
--p-max-len 1600 \
--p-max-ee 2 \
--p-chimera-method consensus \
--p-trunc-len 0
Any suggestions are greatly appreciated—thanks!

forward-seven-number-summaries.tsv (162.4 KB)
per-sample-fastq-counts.tsv (478 Bytes)
dada2 denoise-ccs result , denoising-stats file is shown above.

[Rainjie]

Hi, I'm sorry to bother you, but I really need your help.
I’m using dada2 denoise-ccs . Given the expected amplicon length of 16s rRNA sequence, I’m currently using:
--p-min-len 1000 \
--p-max-len 1600 \
--p-max-ee 2 \
--p-chimera-method consensus \
--p-trunc-len 0
Any suggestions are greatly appreciated—thanks!

forward-seven-number-summaries.tsv (162.4 KB)
per-sample-fastq-counts.tsv (478 Bytes)

[cherman2 (Chloe Herman)]

Hi @Rainjie,
Thank you for your patience!

I am not very familiar with dada2 with Pacbio sequences. What I notice when looking at your post-dada2 stats is that after filtering, we lose the majority of the sequences (~85 to ~60). I am wondering if messing with your --p-min-len may help?

Have you tried a lower min sequence length?

[Rainjie]

@cherman2

if messing with your

Thanks a lot! We’ll try it out first and see how it goes. I’ll follow up in about two days (after the weekend)

[Rainjie]

Hi @cherman2
Thanks for your suggestion!
Despite repeated attempts, we were unable to further improve this outcome. Under the same raw data and identical parameter settings, except expanding the allowable sequence-length range to 800–1600 bp did not alter the results; approximately 40% of sequences were still removed during filtering. So, that's that, we therefore proceeded with the subsequent analyses.

All the best
Rainjie