Question about Discarded field in SAMPLES_QUANTIFICATION_SUMMARY.txt file for CRISPRessoPooled

[mbiokyle29]

We are starting to run larger pooled samples using the CRISPRessoPooled command and I had a question regarding the SAMPLES_QUANTIFICATION_SUMMARY.txt file and the Discarded field in particular. My understanding of the pooled mode is that it is a 2 step processing:

• Run FLASH to merge into read pairs. Read dropout can occur here and is recorded in the top level CRISPResso_mapping_statistics.txt file (READS IN INPUTS vs READS AFTER PREPROCESSING).
• Global demultiplex or alignment to assign reads to their best amplicons. Read dropout can occur here and is also reporting in the top level mapping stats file as READS ALIGNED.
• Then for each amplicon run "normal" CRISPResso on that amplicon with the set of reads that aligned to it. Read dropout can occur here it seems (my guess is that the alignment rules are more strict here). As far as I can tell, this dropout is recorded in the per amplicon CRISPResso_mapping_statistics.txt file, by comparing READS AFTER PREPROCESSING to READS ALIGNED.

Assuming my understanding above is correct, I have the following questions:

• On a per amplicon basis, is it possible that read dropout can occur as part of any other pre or post processing steps besides alignment? The per amplicon CRISPResso_mapping_statistics.txt file has READS IN INPUTS and READS AFTER PREPROCESSING fields, which suggests that its possible but my assumption is that this is just a side effect of re-using the file format from a normal CRISPresso run.
• As far as I can tell, the Discarded field in CRISPResso_mapping_statistics.txt at the top level is always 0 (or null if the amplicon got no reads). It is even 0 in cases where the alignment rate was not 100% (i.e. READS ALIGNED < READS AFTER PREPROCESSING for that amplicon). So if these reads are not considered Discarded, I am curious what discarded reads are and what determines that they should be discarded.

[kclem]

HI @mbiokyle29,

The Discarded field refers to reads that are discarded because the --discard_indel_read flag is set and the read contained an insertion or deletion. However, if this flag is not set, this field will always be 0.

Right - CRISPRessoPooled is a two-step process 1) Demultiplexing to amplicons and 2) Amplicon-specific analysis. The numbers and assignments of reads are contained in two top-level files, one for each step.

1. Demultiplexing to amplicons. First, the reads are trimmed and merged, then aligned to the target using bowtie2. The number of reads at different steps of this process are reported in MAPPING_STATISTICS.txt

• READS IN INPUTS # Number of total reads in CRISPRessoPooled input
• READS AFTER PREPROCESSING # Number of total reads after preprocessing (read trimming and r1/r2 merging)
• READS ALIGNED # Number of reads aligned to the target using bowtie2. In this case, the target can be the whole genome (if running with -x) or individual amplicons (if not running with -x). This number will also include reads correctly aligned to the genome, but not at a site of analysis in the next step. These reads are demultiplexed to respective locations and analyzed in the next step using CRISPResso.

2. Amplicon-specific analysis. Reads assigned to each amplicon location are analyzed using CRISPResso. The numbers and classifications of reads are reported in SAMPLES_QUANTIFICATION_SUMMARY.txt

Columns are:

• Reads_total # Number of reads assigned to this amplicon after demultiplexing
• Reads_aligned # Number of reads successfully aligned using CRISPResso. Reads could be filtered at this step if quality filters are present, and alignments are subject to the read alignment parameter cutoffs (e.g. --default_min_aln_score which by default requires 60% homology).
• Unmodified # Number of aligned reads with no mutations in the quantification window.
• Modified # Number of aligned reads with a mutation in the quantification window
• Discarded # Number of aligned reads discarded because they contain an indel and the --discard_indel_read flag is set

This file aggregates and reports the "READS IN INPUTS" And "READS ALIGNED" values from the "CRISPResso_mapping_statistics.txt" file in each CRISPResso sub-run. These sub-run files also contain the number of reads filtered by preprocessing ("READS AFTER PREPROCESSING") if that is needed. You can check out this discussion on how to troubleshoot reads not aligning on a target amplicon basis.

Does that make sense?

[mbiokyle29]

@kclem - thanks for clarifying all of that. Sounds like if we know --discard_indel_read is not set to true we can safely ignore discarded, which is great. My goal here is to track (and ultimately collect) all dropped out reads. This extra info is very helpful.

[kclem]

Sounds good - if you use the --keep_intermediate flag you will see the flash r1/r2 merging output, as well as the bowtie alignment bam file that was used previous to the demultiplexing.