squigualiser

Move table

Nanopore basecallers output move arrays in SAM/BAM format. The important fields are listed below.

  1. read_id
  2. basecalled fastq sequence length
  3. basecalled fastq sequence
  4. raw signal length in ns tag
  5. raw signal trim offset in ts tag
  6. move table in mv tag
  7. stride used in the neural network (down sampling factor) in mv tag

An example move table looks like the following,

How the auxiliary field is stored in SAM format -> mv:B:c:5,1,1,0,1,0,0,0,1,0,1,0,1,0,0,0,1,0,1,0,1,1,0,1,0,1,0,1,1,1,1,…
Stride (always the first integer) -> 5
The actual move array (the rest) -> 1,1,0,1,0,0,0,1,0,1,0,1,0,0,0,1,0,1,0,1,1,0,1,0,1,0,1,1,1,1,…

The number of ones (1) in the actual move array equals to the fastq sequence length. According to the above example the first move corresponds with 1 x stride signal points. The second move corresponds with 2 x stride signal points. The third with 4 x stride, the fourth with 2 x stride and so on (see illustration below).

image

Different models have different strides. Strides from some of the guppy models are listed below.

Model Stride Guppy version
dna_r10.4.1_e8.2_400bps_fast.cfg 5 v6.3.7
dna_r10.4.1_e8.2_400bps_hac.cfg 5 v6.3.7
dna_r10.4.1_e8.2_400bps_sup.cfg 5 v6.3.7
dna_r9.4.1_450bps_fast_prom.cfg 5 v6.3.7
dna_r9.4.1_450bps_hac_prom.cfg 5 v6.3.7
dna_r9.4.1_450bps_sup_prom.cfg 5 v6.3.7
rna_r9.4.1_70bps_fast_prom.cfg 12 v6.3.7
rna_r9.4.1_70bps_hac_prom.cfg 10 v6.3.7
     
dna_r10.4.1_e8.2_400bps_fast.cfg 5 v6.5.7
dna_r10.4.1_e8.2_400bps_fast_prom.cfg 5 v6.5.7
dna_r10.4.1_e8.2_400bps_hac.cfg 5 v6.5.7
dna_r10.4.1_e8.2_400bps_hac_prom.cfg 5 v6.5.7
dna_r10.4.1_e8.2_400bps_sup.cfg 5 v6.5.7
dna_r9.4.1_450bps_fast.cfg 5 v6.5.7
dna_r9.4.1_450bps_fast_prom.cfg 5 v6.5.7
dna_r9.4.1_450bps_hac.cfg 5 v6.5.7
dna_r9.4.1_450bps_hac_prom.cfg 5 v6.5.7
dna_r9.4.1_450bps_sup.cfg 5 v6.5.7
dna_r9.4.1_450bps_sup_prom.cfg 5 v6.5.7
rna_r9.4.1_70bps_fast_prom.cfg 12 v6.5.7
rna_r9.4.1_70bps_fast.cfg 12 v6.5.7
rna_r9.4.1_70bps_hac.cfg 10 v6.5.7
rna_r9.4.1_70bps_hac_prom.cfg 10 v6.5.7

As of dorado_v0.4.0 stride is used in three stages.

  1. In the process of creating chunks to feed into the NN model (code).
  2. In NN calculations on GPU (have a look at stride in the config.toml file inside a model directory).
  3. To properly stitch the output of each chunk to generate the final fastq sequence (code). This confirms that the NN model is down sampling the raw signal.
This site is open source. Improve this page.