[BUG] sqanti3_filter.py is not filtering according to my custom rules

[dudududu12138]

Is there an existing issue for this?

• I have searched the existing issues

Have you loaded the SQANTI3.env conda environment?

• I have loaded the SQANTI3.env conda environment

Problem description

Hi, thanks for your useful tool. But I met a wired problem recently. I ran sqanti3_qc.py firstly and then ran sqanti3_filter.py to filter out transcripts that don't match my rules.
Below is my custom filtering rules:

{
    "full-splice_match": [
        {
            "perc_A_downstream_TTS":[0,59]
        }
    ],
    "rest":[
        {
            "perc_A_downstream_TTS":[0,59],
            "RTS_stage":"FALSE",
            "all_canonical":"canonical",
            "predicted_NMD":"FALSE"
        },
        {
            "perc_A_downstream_TTS":[0,59],
            "RTS_stage":"FALSE",
            "all_canonical":"canonical",
            "predicted_NMD":"NA"
        },
        {
            "perc_A_downstream_TTS":[0,59],
            "RTS_stage":"FALSE",
            "all_canonical":"NA",
            "predicted_NMD":"FALSE"
        },
        {
            "perc_A_downstream_TTS":[0,59],
            "RTS_stage":"FALSE",
            "all_canonical":"NA",
            "predicted_NMD":"NA"
        }           
    ]
}

Code sample

This is my code:

sqanti3_filter.py rules ${input}_classification.txt \
 -j filtering.json \
 --isoforms ${input}_corrected.fasta    \
 --gtf ${input}_corrected.gtf \
 --faa ${input}_corrected.faa \
 --skip_report \
 -d $output -o syfilter 

Error

Some transcripts fulfil my criteria and should be recognised as isoform, but are filtered out by sqanti3. Here is an example:
This isoform was filtered because of out all_canonical.

But I checked classification.txt and junxtions.txt, all junctions in this isoform are canonical.
Below are the information of this isoform in the two files:

• classification.txt:
  

• junctions.txt:
  

So why did such isoforms be filtered out?

Anything else?

No response

[dudududu12138]

Hi, I have checked your source code utilities/filter/rules_filter_functions.R and I found the reason. You checked whether the column of the isoform_info was NA firstly (if (! is.na(isoform_info[rules[i, "column"]]))). If it was NA, it was classified as Artifact. But in my custom rules, I defined some NA rules. Such as the all_canonical column could be NA so that the mono-exon isoform can be checked. Also the predicted-NMD column. I think your codes prefer to coding transcripts while ncRNA will be incorrectly filtered out. Below is the section of your code:

if (! is.na(isoform_info[rules[i, "column"]])){
          if (rules[i, "type"] == "Min_Threshold"){
            if (as.numeric(isoform_info[rules[i, "column"]]) < as.numeric(rules[i, "rule"])){
              is_isoform=FALSE
              break
            }
          }else if (rules[i, "type"] == "Max_Threshold"){
            if (as.numeric(isoform_info[rules[i, "column"]]) > as.numeric(rules[i, "rule"])){
              is_isoform=FALSE
              break
            }
          }else if (rules[i, "type"] == "Category"){
            cat_rules <- rules[rules$column == rules[i, "column"], ]
            if ( ! tolower(isoform_info[rules[i, "column"]]) %in% cat_rules[,"rule"]){
              is_isoform=FALSE
              break
            }
          }
        }else{
         is_isoform=FALSE
         break
        }