Seeking Advice on Customized Illumina Sequencing for Low-Diversity CRISPRi Libraries

[edelvinm]

Hello community,

Overview of the Experiment:

I am working with a CRISPRi plasmid library to identify bacterial genome-wide fitness effects of silenced genes under antibiotic exposure. My library contains 10,000 sgRNAs, each 20 nucleotides long. Sequencing was provided by a company using the standard Illumina protocol.

For sample preparation, I custom barcoded 8 samples (4 nucleotides long for in-house demultiplexing) and pooled them before sequencing. The service provider ligated adapters for routine sequencing on the NextSeq 500 Illumina platform, generating 12 million reads in total, with an average depth of at least 250 reads per sgRNA (PE150).

Issue Encountered:

When analyzing the results, I found that about 95% of the library was lost, leading to low library diversity. Even in the initial untreated samples, only about 5% of the library was observed.
Upon reviewing the literature, I discovered custom Illumina protocols designed to address low library diversity issues. These protocols suggest:

1. Research Article 1:

- "A customized Illumina sequencing method was designed to avoid problems arising from low library diversity when sequencing PCR products. The first 2 cycles that read bases common to all clusters were set as dark cycles, followed by 20 cycles to read the guide RNA."

[Reference]

2. Protocol Article 2:

- "The custom sequencing recipe is designed to avoid issues caused by low-diversity sequences of amplicons. The first 54 cycles in the run are for sequencing common sequences of the amplicons and are set as dark cycles (i.e., the chemical reactions occur, but no imaging). The following 20 cycles are to sequence the diversified base-pairing region of sgRNAs."
[Reference]

Next Steps:

I plan to contact companies capable of implementing this customized cycle setup for Illumina sequencing. Since this will be my first experience with such a sequencing method, I would like to understand the reasons behind these protocols better before reaching out. For what reason would the custom cycle recover the remaining library during sequencing? The literature doesn’t clearly explain the rationale for these specific sequencing methods.
Any insights or experiences you could share on this sequencing approach and the reasons behind these custom protocols would be greatly appreciated!
Thank you!

[ADoetsch]

The problem with low diversity samples is that in some cycles nearly all of the clusters show the same base resulting in at least one of the color channels becoming very dark. This can mess up the base calling algorithm and subsequently lead to loss of whole sequences due to quality filters.
The common recommendation is to use higher amounts of PhiX library to introduce background variation and thus balance out such low-diversity loci (see, e.g. https://knowledge.illumina.com/instr...list/000002882).

Do you know, if your provider used PhiX and if so, how much? We are doing regular runs of 16S-rRNA on the MiSeq platform. PhiX concentration should be increased (>= 20%) to get reliable quality.

I'm not familiar with setting up dark cycles (don't even know, if the MiSeq can do this) but it would also solve the problem if you known exactly where your low diversity sequences are. I expect it to still fail if there are few bases in between with low diversity. PhiX should solve this better, although you'll sacrifice a part of your data for it.

[edelvinm]

Thank you for your response, it was very helpful. Checking your link and reading additional information about "nucleotide diversity" and "PhiX control" (link1; link2) clarified the problem for me. I also compared my FastQC report (attached) with the Percent Base by Cycle plot for low library samples on the Illumina website and noticed the similarities (link3).



Regarding your question, I am not sure if the service provider used PhiX. If my samples were mixed with other diverse samples from other customers during sequencing, would this add enough background variation to the base composition to mitigate the low diversity issue, similar to what PhiX does? Additionally, my amplicon size is 213 bp long, with the only diverse 20nt region starting from the 38th base position.

Thank you again for your insights!

[ADoetsch]

PhiX reads are usually not exported with your fastq files, so the analysis with FASTQC will not tell you, if PhiX was included. You'd need to have a look at the QC images of the platform itself, which you likely don't have access to. You need to ask the service provider.

What the graph does tell you that you have indeed loci with low diversity, since there are several locations, where at least one of the bases is nearly zero. This will cause quality issues, if there is no other DNA in the run that compensates for this.