-
Are too many reads being rejected?
You can determine the number of reads that are being rejecting due to adaptive sampling via two methods:
- During the run: using the read length histogram in MinKNOW by ticking the box Split by read end reason. If you are running in a mode where a sufficient number of reads should be rejected to produce a peak of short reads, there should be a bimodal distribution with a peak at a base-count much lower than what you expect from the sample. The reads rejected by read until will have the label Data Service Unblock Mux Change.
- After the run: using the
adaptive_sampling.csv
. This file will contain information about the final decision that was taken by adaptive sampling for each read.
-
Am I not getting a sufficient enrichment from my sample?
- Are your reads long enough? Short reads will spend most of the time in the decision phase, which means the software will select for a low fraction of bases.
- Is your input amount sufficient? Adaptive sampling increases the number of transitions from strand to pore to allow recapture. Therefore, having a low input amount with an associated long time to capture the strands can exacerbate this.
- Does your output decrease quickly over time? You may need to wash your flow cell using the Flow Cell Wash Kit to recover some channels, then reload more of your sample.
-
Am I not getting my entire region of interest enriched?
If you are using a .bed file to perform enrichment, you may need to extend the regions of interest to include more bases at the beginning and end, where you are not seeing sufficient enrichment.
-
How many bases are read before accepting or rejecting a strand?
For enrichment, 200 bases are required before a decision is made. However in practice, due to MinKNOW read detection the number is typically ~450 bases for R9.4.1 chemistry.
For depletion, the minimum number of bases required is ~450, but up to 4000 bases can be read before a strand is ejected.
-
What happens to rejected reads?
If a strand has already translocated through the pore and its sequence has been aligned to make the rejection decision, the read is output to the
fast_q pass
folder if it passes the Q-score filter, or to thefast_q fail
folder if it falls below the Q-score filter.Rejected reads will not be sequenced again. Rejected reads are included in the total outputs reported by MinKNOW, however as these reads are typically short in a good adaptive sampling run, this should make a minor difference to the enriched output value.
-
Should I use a reference index or reference sequence?
Either is suitable, however if using a reference sequence, the run setup may take longer as MinKNOW builds the reference index (minimap2 file). You can use MinKNOW to build the minimap2 index ahead of run setup if required - see the MinKNOW protocol for details.
-
Is adaptive sampling impacted by barcodes, especially if I have custom barcodes that increase the length of the sequence before my material? Can I de-multiplex while running adaptive sampling?
Adaptive sampling should not be impacted by barcodes, and demutiplexing is possible in adaptive sampling mode.
-
Should I use FASTA alone or also a .bed file? Does this make a difference to performance in different circumstances?
Either option will work, however inputting both a FASTA and .bed file means that reads can be potentially aligned to either file, and therefore be easier to reject.
-
If I have a GPU on my computer for use with the MinION, and select HAC basecalling while running adaptive sampling, can this lead to issues in the performance of adaptive sampling or basecalling?
Yes - both adaptive sampling and basecalling use the GPU, so this can lead to performance issues such as adaptive sampling falling behind and reads not being rejected and fully sequenced.