search

LooseLab / readfish

CLI tool for flexible and fast adaptive sampling on ONT sequencers
https://looselab.github.io/readfish/
GNU General Public License v3.0
163 stars 31 forks source link

comparison between MinKNOW and Readfish #357

Closed ylang0825 closed 1 month ago

ylang0825 commented 1 month ago

Hello colleagues, Recently we conducted two enriment experiments. For the first one, we used MinKNOW to enrich a target reference from 1 to 256 channels, and the remaining 256 to 512 channels served as control. For the second one, we used Readfish to enrich a same target reference, and we also set a Readfish group and a control group in the toml file. Our data analysis indicated that the adaptive sampling groups of MinKNOW and Readfish procuded similar results such as base number and target coverage depth. However, the base number of the control group in the MinKNOW trial was 1.5 times that of the control group in the Readfish trial. This bothered us a lot. We noticed that Readfish split the channels to different groups based on physical positions of channels, which is different from MinKNOW. We would like to know whether this could be a factor influencing the output of control groups. And whether we can select some ranges of channels as control like MinKNOW? If so, how can we achieve that. If not, could you please explain why. Thank you very much!

github-actions[bot] commented 1 month ago

Thank you for your issue. Give us a little time to review it.

PS. You might want to check the FAQ if you haven't done so already.

This is an automated reply, generated by FAQtory

mattloose commented 1 month ago

Hi

Can you provide more information on your experimental design here.

Did you flush and reload the flowcell between running the Minknow and readfish controls? If you used different flow cells were they matched with respect to the number of pores on each?

If you used the same flowcell for the two experiments then a change in absolute yield from the control portion is to be expected.

Thanks.

ylang0825 commented 1 month ago

Thanks for your quick response. We used two new flow cells. One for MinKNOW and control, and the other for Readfish and control. Both flow cells passed the quality control and showed similar active channels. We compared the results after 48 hours, and we did not flush and reload the flowcell. Besides, we analyzed the drop of active channels during two experiments. The active channels dropped similarly in the adaptive sampling groups for MinKNOW and Readfish, but the active channels dropped slower in the control group of Readfish relative to the control group of MinKNOW. By the way, I realized that I have stated wrong in the above question. The fact is that the base number of the control group in the Readfish trial was 1.5 times that of the control group in the MinKNOW trial, consistent with the drop of active channels. We speculated whether this resulted from different partition of channels in control groups? Thank you!

mattloose commented 1 month ago

The key issue is not the number of active channels, but the number of pores on the flow cell. If you had more pores on the Readfish flowcell you would expect more data (just double checking that you are using pore count and not channel count).

When Readfish splits the flowcell we get two physical blocks which are colocalised. I believe when MinKNOW splits a flowcell the adaptive and non adaptive channels are mixed in with one another. This may have an effect on performance but we haven't looked at it extensively.