Our latest Kit 14 chemistry is an upgrade across our product range to enable highly accurate and tuneable sequencing.

Upgrades include:

• New E8.2 motor protein in our V14 sequencing kits (e.g SQK-LSK114)
• New R10.4.1 flow cells
• Basecalling upgrades for simplex or duplex data

Kit 14 builds on our previous chemistry upgrades to combine high output with Q20+ accuracy. Duplex rates can also be achieved to further improve accuracy to Q30+.

Note: For highest sequencing accuracies, ensure the super-accurate basecaller (SUP) is selected for your sequencing parameters during run set-up on MinKNOW.

Figure 1. Example data expected to be achieved using the Ligation Sequencing Kit V14 (SQK-LSK114) on a GridION Flow Cell.

Our Kit 14 chemistry has been developed to reach sequencing accuracies of ~Q30 with duplex data and SUP basecalling. Duplex basecalling is the reading of both strands for even higher accuracy compared to our previous kits.

We offer two options for basecalling Kit 14 data with sequencing accuracies of 99% and above (Q20+):

• Simplex basecalling: The sequencing of a single strand. The template DNA strand passes through the nanopore and is basecalled. This is performed on MinKNOW.
• Duplex basecalling: The sequencing of both strands. The complement strand is read immediately after the template strand and the consensus basecall for both strands leads to a further increase in accuracy. This is performed on either Dorado.

Duplex reads do occur at a lower level naturally. However, our Kit 14 chemistry include upgrades in the motor protein and pore in the flow cells, which in combination work to increase second strand capture to improve duplex rates. For more information on basecalling duplex data, please see the "Basecalling Kit 14 duplex data" section.

Below, we illustrate duplex sequencing at the pore level (Fig. 2) and we demonstrate what data can be achieved with duplex sequencing (Fig. 3).

Figure 2. The diagram above demonstrates how duplex reads are sequenced on the flow cell, the first strand is the template strand and the second is the complement strand.

Figure 3. This graph illustrates percentage of data in a duplex pair. Duplex data is presented for both template and complement strands, therefore, for this example, duplex rate is 21.4%.

Simplex data is generated when a single strand of DNA is sequenced. It happens at a much higher volume than duplex sequencing, meaning higher outputs can be generated.

For experiments that require high data output, we recommend loading your flow cell with a high amount of library (35-50 fmol) to ensure high pore occupancy of >95% is reached to maximise data output. If your sequencing run starts with a lower pore occupancy after 1 hour of starting, data output may be negatively impacted.

Pore occupancy is the percentage of pores that are sequencing at any time out of those available for sequencing. This is displayed in the pore activity plot in MinKNOW during an experiment. During the first few minutes of sequencing, pore occupancy can increase as the library tethers to the membrane. Therefore, we recommend reviewing pore occupancy ~1 hour into a sequencing run.

To calculate pore occupancy:

Simplex basecalling is where the single template DNA strand passes through the nanopore and is basecalled with sequencing accuracies of up to 99% when using our super-accurate basecaller (SUP). We recommend basecalling Kit 14 simplex data on MinKNOW.

For optimal performance on our devices, please see the specifications below:

• 64 bit Linux or Windows 10
• Intel i7, i9, Xeon, or better processor
• At least 16 GB of RAM
• An NVIDIA GPU, at least RTX 2070 or better, with at least 16 GB of GPU memory
• At least 1 TB SSD

Make sure you are using the most recent version of MinKNOW.

Below are our basecaller and output recommendations for setting up a Kit 14 sequencing run on MinKNOW to generate POD5 files to use as input for duplex basecalling on Dorado.

Duplex basecalling is performed on Dorado after simplex basecalling is completed on MinKNOW. Duplex basecalling is where the complement strand is read immediately after the template strand and the consensus basecall for both strands leads to a further increase in accuracy to ~Q30 with our super-accurate basecaller (SUP).

To generate high duplex output, it is important to follow the library preparation protocol to ensure successful ligation of sequencing adapters onto both ends of the DNA strands. It is also important to follow the flow cell loading recommendation of 10-20 fmols for optimal duplex output to ensure the flow cell is not under- or overloaded:

• Underloading results in lower capture rate and less overall output.
• Overloading adds competition of other strands around the nanopore and reduces the rate of duplex read capture.

For calculating how much DNA to load onto your flow cell, please use the table below for guidance. For more information on DNA input, please see the '[Input DNA/RNA QC](https://community.nanoporetech.com/docs/prepare/library_prep_protocols/input-dna-rna-qc)' document.

Mass	Molarity if fragment length = 2 kb	Molarity if fragment length = 8 kb	Molarity if fragment length = 50 kb
1 μg	770 fmol	193 fmol	31 fmol
500 ng	395 fmol	96 fmol	15 fmol
400 ng	308 fmol	77 fmol	12 fmol
200 ng	154 fmol	39 fmol	6.2 fmol
100 ng	77 fmol	19 fmol	3.1 fmol
75 ng	61 fmol	15 fmol	2.4 fmol
50 ng	40 fmol	10 fmol	1.6 fmol
30 ng	23 fmol	5.8 fmol	0.9 fmol
20 ng	16 fmol	4 fmol	0.6 fmol
15 ng	12 fmol	3 fmol	0.4 fmol
10 ng	7.7 fmol	1.9 fmol	0.3 fmol

To basecall duplex data, we now offer duplex read basecalling in Dorado (recommended) and Guppy, where the template and complement strands of a read can have their basecall data combined to provide a more accurate sequence. For more information, please see our [Dorado page on Github](https://github.com/nanoporetech/dorado).

Please note, to generate data of the highest accuracy, ensure data is basecalled using our super-accurate basecaller (SUP).

1. Set up sequencing parameters in MinKNOW to perform simplex basecalling as described in "Basecalling Kit 14 simplex data". a. Basecall using the Fast basecaller. b. Output .POD5 files.

2. Using Dorado, re-basecall your simplex data with the following command to output simplex and duplex reads using the super-accurate (SUP) basecalling model. Other basecalling models can be used, as listed on the Dorado Github page.

$ dorado duplex dna_r10.4.1_e8.2_400bps_sup@v4.1.0 pod5s/ > duplex.bam

Dorado is a high-performance basecaller which is used to perform duplex basecalling. For further information about Dorado, please see the Dorado Github page.

Note: When running Dorado, we recommend stopping other basecalling for the best performance by maximising memory available to Dorado. This can be stopped and restarted when Dorado has finished via the GUI on MinKNOW.

Guppy may also be used to duplex basecall.

Duplex data can be presented as two values:

• Duplex rate as a percentage of bases or reads in a duplex pair
• Duplex output as Gb of data in a duplex pair

To determine total percentage of duplex reads, we calculate as follows:

= ((template + complement)/total) * 100

Figure 4. This graph illustrates the duplex reads presented as template strand and complement strand, explaining the need for the calculation to combine the reads to calculate the total percentage. In this example, duplex rate would be 21.4%.

Read-splitting is a default parameter in MinKNOW to split the informatic chimeras, as explained below. In our Kit 14 chemistry, we have observed increased follow-on rates, and with increased follow-on rates (the rate of the complement strand entering the pore directly after the template strand has passed through), we have observed a higher rate of concatemerisation compared to the Ligation Sequencing Kit (SQK-LSK110). We are classifying these reads as 'informatic chimeras' as they are not physically joined during the library preparation process.

With SQK-LSK110, we typically observe <2% concatemerisation and at this rate, it typically does not affect downstream applications. With, for example SQK-LSK112, we have observed a rate as high as 10%. Both MinKNOW (v21.11 and higher) and stand-alone Guppy (v5.1 and higher) now offer the option of splitting these reads.

As you can see from the following example, the majority of the informatic chimeras (yellow) are removed after splitting for a human (native) and E. coli (PCR) sample.

It is important to note that the read splitting function is not designed to split reads that are incorrectly ligated together during sample preparation. While these make up a small percentage of reads, users should take care with ligation steps to follow the protocol carefully to reduce the chance of creating them.

Flye is recommended as an assembly tool for Kit 14 genome assembly (https://github.com/fenderglass/Flye).

We have observed that assembling haplotypes separately significantly improves genome contiguity, although each assembly only uses half the data.

1. Configuring the command line parameter --min-overlap 10000 should deliver a modest improvement in assembly contiguity when using libraries optimised for read length.
2. It is recommended that the --nano-corr parameter is set (to specify that the sequences are "corrected"). This provides a significant improvement to assembly NG50 compared to when the --nano-raw (uncorrected sequence) setting is used. We have observed NG50 increases from 58 Mb to 67 Mb for collapsed assemblies, when assembling both haplotypes at once.
3. We typically adjust the "asm_corrected_reads.cfg file in the flye/config/bin_cfg/ folder to increase haplotype-specific assembly NG50s and to remove any major misjoins.

a. enable homopolymer compressed scoring (hpc_scoring_on = 1) b. increase the minimizer_window to 10 c. decrease the repeat_graph_ovlp_divergence to 0.005 increases haplotype-specific assembly NG50s to 84 Mb/84 Mb and removes all major misjoins