Training datasets for MASSEQ model

[Jiayi-Wang-Joey]

Dear developers,

Thanks for the amazing work.
I have a quick question: which datasets you used for training MASSEQ model?

Maybe you already noted the information somewhere, but I couldn't find it. Could you remind me here?

Thanks again!

Kind regards,
Jiayi

[kishwarshafin]

Hi Jiayi,

The model was trained on a combination of PacBio full-length RNA sequencing datasets mapped to GRCh38, capturing both the original MAS-Seq chemistry (mid-2023) and the higher-throughput Kinnex chemistry (late 2023).

Specifically, the training data utilizes standard Genome in a Bottle (GIAB) reference cell lines that have been processed through the standard Iso-Seq pipeline (demultiplexed with lima and trimmed via isoseq refine). These are the samples used:

• HG002
• HG004
• HG005
• NA26105
• NA27730

[Jiayi-Wang-Joey]

Hi, thanks for reply! This sounds you used all GIAB datasets for training. Which datasets did you use for testing?

[kishwarshafin]

@Jiayi-Wang-Joey in all of our training for all models HG003 is always held out and chr20 is always held out.