Selection of True Negatives

[mpizzagalli777]

Hi,
Thanks so much for creating this powerful tool. I was wondering if someone would be able to provide some guidance for how to define true negatives for my dataset in order to do the machine learning based filtering.

For context, I have three sets of nanopore sequencing data from which I would like to generate a high confidence transcriptome. I have performed flair align, correct and collapse to merge together the three datasets. SQANTI QC shows that I have a relatively good transcriptome (I think), with the following breakdown of transcript identities:

Before checking these numbers, I ran sqanti ml filter but due to the fact that I have so few NNI transcripts, the filter is unable to generate the training data. I was wondering if you had any suggestions on how to select a list of TNs? And also if you have any idea why I have so few TNs in general. Is this a situation where I will not be able to use the ml filter and will have to rely on the rules based filter instead? From the paper it seemed the ml filter was more powerful so I was hoping to use this approach.

Please let me know if you would need any additional information and thanks in advance!

[mpizzagalli777]

Also, for some reason the GB2_merged_QC_TN_list.txt file is empty, even though as you can see above I have many RM transcripts. Is this expected due to the error stemming from the lack of NNI transcripts?