Options for demultiplexing libraries with different index lengths in the same sequencing run

Mixing libraries of different types (such as single and dual indexes, or different index lengths) are not recommended due to different clustering efficiencies and potential issues with demultiplexing. Illumina supports running different library types in different lanes of flow cells that support individual lane loading but does not recommend mixing library types within the same lane of any flow cell.

The following reviews information on how to set up and demultiplex runs with mixed library types depending on the software used, if necessary.

Run setup considerations

• Illumina Experiment Manager does not allow for different indexing lengths for sample sheet creation. Generate the sample sheet for the libraries with the longest indexing scheme, then manually edit the CSV to include the other libraries as appropriate.

• Local Run Manager

  • Local Run Manager v1 for MiniSeq does not allow for mixed index types. Local Run Manager v1 only uses index kit selections (there is no free-text option), and while custom kits are supported, custom kits cannot contain Ns (wildcard bases). Importing run data into a run setup where the index length does not match the RunInfo.xml file will result in an error.

  • Local Run Manager v2 and v3 allow for mixed index types if the library type is set as Custom, and the entered lengths for each sample match. Ns can be used to pad the shorter length indexes.

• BaseSpace Preptab does not allow for mixed index types and runs set up through Preptab cannot be requeued with a different kit. Set up the run with the longest indexing type and see demultiplexing options below.

• BaseSpace Run Planning does not support N-padding. Depending on the instrument type runs with different index lengths need to be set up in different ways:

  • NovaSeq X/X Plus: Set up the run initially for the longest index lengths. Using the multi-analysis feature set up a different analysis for each library type and the OverrideCycles setting for BCL Convert will be used as appropriate for each library type.

  • NextSeq 1000/2000 and NovaSeq 6000/6000Dx: Multi-analysis is not supported. Set up runs for the longest lengths and either perform multiple rounds of demultiplexing or analyze with BCL Convert v4.1.5 or higher using per-sample settings (see below).

Demultiplexing options

• bcl2fastq:

  • v2.20: - If library types are consistent per lane (single indexes in one lane, dual in a different lane, etc.), create a single sample sheet with the indexes as-is. For any single indexed libraries leave the index2 column blank. This version of the software will auto-adjust the index cycle usage per lane. N padding is not supported. - If library types are mixed within lanes then create multiple sample sheets for each index type and perform multiple rounds of demultiplexing. This version of the software will auto-adjust the index cycle usage based on the sample sheet as long as index length usage is consistent (do not mix 6 and 8 bases in an analysis, for example). N padding is not supported.

  • v1.8: - Mixed index types are not supported, even in different lanes. Create multiple sample sheets for each index type and perform multiple rounds of demultiplexing. Set the --use-bases-mask option to appropriately mask out the run cycles for each analysis as necessary. N padding is not supported.

• BaseSpace Sequence Hub (BSSH): BSSH uses bcl2fastq 2.20 for FASTQ generation by default for most instrument types (NextSeq 1000/2000 and NovaSeq X/X Plus use BCL Convert, while NovaSeq 6000/6000Dx users can set up their runs for either software). See the rules for bcl2fastq 2.20, above.

  • For Preptab runs with mixed index types, customers will need to either download the run data and reanalyze locally with bcl2fastq or Local Run Manager (LRM), or use the BCL Convert app with a v2 sample sheet in BSSH.

• BCL Convert: As of version 4.1.5, per-sample settings are supported, allowing for sample-specific usage of OverrideCycles. See Upgrading from bcl2fastq to BCL Convert for more details on per-sample settings.

  • For earlier versions of BCL Convert, the software does not support per-sample settings or per-lane masking like bcl2fastq 2.20. Create multiple sample sheets for each index type and perform multiple rounds of analysis. Use the OverrideCycles sample sheet option to set the index base mask as appropriate for each analysis. N padding is not supported.

  • BaseSpace users can upload sample sheets and run the BCL Convert app to reanalyze their run data.

• MiSeq Reporter (MSR) and Local Run Manager (LRM): Most workflows and modules support N padding, where N is interpreted as any base. Use Ns to make the different index types a consistent length. Runs cannot be requeued using an index length that differs from the RunInfo.xml file.

  • This applies to workflows and modules that use BuildFastq, which is most of the analysis workflow options, including fastq generation (FastQ Only). The following modules use bcl2fastq and do not support N padding or multiple rounds of analyses: DNA Amplicon, RNA Amplicon, RNA Fusion, TruSight Tumor 15

  • Runs set up in MSR or LRM with Ns will demultiplex correctly locally, but not if uploaded to BSSH. For uploaded runs follow the bcl2fastq 2.20 rules and create multiple sample sheets for multiple rounds of analysis.

For any feedback or questions regarding this article (Illumina Knowledge Article #5982), contact Illumina Technical Support [e-mail].