--targeted analysis

[dcarrillox]

Description of feature

Hi, we are currently working with a targeted methylation protocol and thought it would be great if the methylseq pipeline could handle this scenario. Me and my colleague @luigilamparelli will be working on this during the nf-core hackathon 2025 (slack). Here are the main points we plan to implement:

• --targeted parameter: Currently, all detected methylation signals are included in the final results. However, some enrichment protocols target specific genomic regions. To address this, we propose adding a --targeted parameter that allows filtering results to retain only on-target methylation signals. The input would be a BED file specifying the target regions.

• Enrichment metrics: Building on this idea, we propose adding a job to calculate on-target vs. off-target fractions of aligned reads, along with other enrichment metrics. A tool like CollectHsMetrics from Picard could be used for this purpose.

We'll work on this and let you know how it goes!

[bounlu]

How different will the methylation metrics be than running the current pipeline as it is on a targeted data?

[dcarrillox]

For targeted data, and with the pipeline as it is right now, methylation results also account for the off-target, ie. reads that align to other regions than the intended ones. The approach we followed is filtering the bedGraph from bismark/methyldackel with the target BED file.

Only Bismark is available for MultiQC, which gets the metrics from the log file. Since we filter after running bismark, the log files don't change and the multiqc results will still be "whole genome".

[bounlu]

Considering the off-target data is usually about 10-15% for a normal protocol, I am curios to know to what extent the overall methylation patterns will change. It will be interesting to see the difference once you implement this.