This repository contains predictive kmer models for development use. Each folder contains a set of template and complement models and is named in the format

<pore>_<bias_voltage>_<speed>_<kmer_length>[_RNA]

e.g. r9.2_180mv_250bps_6mer

There will usually be one template model and two complement models. The two complement models correspond to "nonuplifted" and "uplifted" populations ("pop1" and "pop2" respectively).

Each model is a tab-delimited text file containing seven columns. These models assume that the current levels for each kmer can be approximated using a Gaussian distribution and the corresponding signal noise can be approximated using an inverse Gaussian distribution.

1. kmer The kmer being modelled.
2. level_mean The mean of a Gaussian distribution representing the current observed for this kmer.
3. level_stdv The standard deviation of the above Gaussian distribution of observed currents for this kmer.
4. sd_mean The mean of an inverse Gaussian distribution representing the noise observed for this kmer.
5. sd_stdv The standard deviation of the above inverse Gaussian distribution of noise observed for this kmer.
6. ig_lambda The lambda parameter for the above inverse Gaussian distribution of noise observed for this kmer.
   (so sd_stdv == sqrt(sd_mean ^ 3 / ig_lambda). See Wikipedia for more info.
7. weight Used internally for model training purposes.

Note that RNA kmer models currently have uracil labelled as thymine, so that it looks like DNA. This is in keeping with the present output format of the event table in basecalled fast5 files, though it may change in future.