search

Psy-Fer / deeplexicon

Signal based nanopore RNA demultiplexing with convolutional neural networks
https://psy-fer.github.io/deeplexicon/
MIT License
36 stars 8 forks source link

Support for RNA004? #27

Open Delayed-Gitification opened 11 months ago

Delayed-Gitification commented 11 months ago

Hi,

Just wondering if you have trained any models on barcodes for RNA004? This would be super useful!

Many thanks Oscar

Psy-Fer commented 11 months ago

Hello Oscar,

We did plan on doing this much earlier, however, we were not given early access to the kits/flowcells, and along with the announcement that nanopore were going to release a barcoding method for dRNA, we put this on a very low-priority (see slow5/pod5 for why)

Fortunately, the original group of researchers who created the deeplexicon method, now make up multiple other research groups. I have discussed this briefly with one of my collaborators and they are willing to do the new barcode designs (many more than 4 this time) and sequence them to generate the training data. Our deep learning specialist is also keen to do the training. If all of this happens, I will update the tooling to integrate it all into a fancy new release.

Unfortunately, I don't have a timeline for you, again, with nanopore barcoding methods 2 weeks or 2 years away (who knows, they certainly don't) it's not exactly an enticing project to throw our time and effort behind in place of other projects.

So tl;dr: We will, but it's not our top priority.

As a little side note, we can do some other cool things with the release of many many more barcodes. So there is at least the enticement of some cool down stream method development that can be done.

Cheers, James

Delayed-Gitification commented 11 months ago

Thanks James. Yes, I can see why this is a difficult project to prioritise given the situation with nanopore. I actually contacted a rep to ask when barcoding is likely to be released by them. I doubt they will tell me but I'll let you know if they do!

Best Oscar

enovoa commented 3 months ago

Dear Oscar, we have been training a different algorithm for RNA004 runs. Feel free to email me directly to get early access to alternative software for demuxing RNA004 runs if you're still interested in this. Thanks! Eva

jibrantahir commented 2 months ago

Hi I am catching up on this old question- is there a possibility to get early access to alternative software for demuxing RNA004 with possibly more barcodes- 96? Also another question- if the barcode sequence is present in the sequences from both ends, why cant we just use this information to demultiplex with conventional methods - like cutadapt, many thanks

enovoa commented 2 months ago

Dear @jibrantahir as mentioned above, please feel free to mail me directly to get early access to alternative software for demuxing RNA004 runs. Thanks! Eva