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

supprot for POD5 format?? #30

Open Seongmin-Jang-1165 opened 2 months ago

Seongmin-Jang-1165 commented 2 months ago

hello

I'm planning to multiplex my dRNA seq library referred to this article

but in the issue tab, developer said this tool do not support RNA-004 library.

is there any plan to develop methods for RNA004, or for POD5 data??

Psy-Fer commented 2 months ago

Hey there,

Please see in the other issues, such as https://github.com/Psy-Fer/deeplexicon/issues/27 That Eva has an alternative to offer.

@enovoa could you please provide contact info for them to get in touch?

Cheers,

James

Seongmin-Jang-1165 commented 2 months ago

@Psy-Fer Thanks you

I have another question.

If I convert POD5 raw data to FAST5, will these files be compatible with Deeplexicon?

Also, does the difference in library kit and flow cell chemistry affect the raw signal itself to the extent that it changes the signal patterns?

Could this lead to issues when analyzing with existing tools like Deeplexicon, or is converting the data to FAST5 sufficient to ensure compatibility?

Psy-Fer commented 2 months ago

Hey,

RNA004 signal won't work with deeplexicon. As it was trained with RNA001/2 on R9.4.1 flowcells.

Conversion won't change the data inside.

As far as deeplexicon goes, it is essentially deprecated software due to LIbrary and flowcell changes. Hopefully new methods are released soon.

Cheers,

James

noncodo commented 2 months ago

Has anyone actually tried running it on RNA004 data? The main chemical difference is the motor protein. Since it uses the same pore as RNA002 (albeit it may have been modified slightly), the discriminant signal may nonetheless still be pertinent, as the time domain is normalized in the GASF transformation. I won't vouch for its accuracy, but it might be 'good enough' to demultiplex a fraction of the reads.

Eva's new aolution may be the best alternative for now, but I'm also eager to refresh Deeplexicon 2.0. Patience for an ONT solution is wearing thin...

From: James Ferguson @.> Sent: Thursday, October 3, 2024 9:41:44 a.m. To: Psy-Fer/deeplexicon @.> Cc: Subscribed @.***> Subject: Re: [Psy-Fer/deeplexicon] supprot for POD5 format?? (Issue #30)

Hey,

RNA004 signal won't work with deeplexicon. As it was trained with RNA001/2 on R9.4.1 flowcells.

Conversion won't change the data inside.

As far as deeplexicon goes, it is essentially deprecated software due to LIbrary and flowcell changes. Hopefully new methods are released soon.

Cheers,

James

— Reply to this email directly, view it on GitHubhttps://github.com/Psy-Fer/deeplexicon/issues/30#issuecomment-2391457168, or unsubscribehttps://github.com/notifications/unsubscribe-auth/ABDCR37744XCTWRKWEXPLJTZZU3QRAVCNFSM6AAAAABOYEOIBKVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDGOJRGQ2TOMJWHA. You are receiving this because you are subscribed to this thread.Message ID: @.***>

Psy-Fer commented 2 months ago

Hey Martin,

Good question. I asked nanopore if they needed a multiplexing option for RNA004 for release and they told me they had a solution....just like when they told me they had one for RNA003 in 2019 and we both know how that turned out.

As always I'm happy to redo deeplexicon for RNA004 if someone wants to generate the data.

James