Replace stampy with minimap2

[bricoletc]

@iqbal-lab could you give me access to the cortex repo? so I could in future push code to a minimap2 branch for eg. In the meantime this is for reviewing my changes.

Code

• Remove stampy command line args in run_calls.pl and process_calls.pl, replacing with a single minimap2_bin argument
• instead of stampy index, use ref fasta (passed as --ref_fasta) to place variants in process_calls.pl (minimap2 can index ref on the fly) Previously, --ref_fasta was used to check for misplaced variants. run_calls.pl always uses the option. Now, this always occurs, as --ref_fasta is used to place variants.

Checks

I have validated that I get the same VCF when running py_cortex_api (it wraps cortex's independent workflow) with stampy and with minimap2, on cortex's demo files example1

TODOs

• A systematic check on several datasets- notably, I observed minimap2 MAPQ seems lower than stampy for 50bp reads
• Update cortex manual
• Maybe this code can be removed, as minimap2 could probably map >1000bp sequence even in short-read preset? : https://github.com/iqbal-lab/cortex/blob/6d8e2f9f1984651c253bdd9d748acf6efa54c165/scripts/analyse_variants/process_calls.pl#L2682-L2691

[bricoletc]

Logging here a validation evaluation of using minimap2. Evaluation uses varifier on one staph dataset from martin (with truth genome assembled using pilon), :

I found that the 'good' minimap2 settings (-k9 -w4) give following recall/precision:
  * require mapq > 40: "Precision_edit_dist": 0.99996956, "Recall_edit_dist": 0.50726812
  * require mapq > 0:  "Precision_edit_dist": 0.99998191, "Recall_edit_dist": 0.83881474,
  * previous cortex, using stampy and mapq > 40, gets "Precision_edit_dist": 0.98863584, "Recall_edit_dist": 0.83029847,
on this dataset, minimap2 with mapq>0 does a little better in both metrics.

[bricoletc]

Logging some more validation work using 14 ilmn Plasmodium falciparum samples with matches pacb assemblies (in previous comment, ran on a Staphylococcus aureus dataset):

stampy: recall: 0.4490 precision: 0.9118
minimap2: recall: 0.4443 precision: 0.9258

These are the average of metrics computed using varifier. THis confirms the change works fine.

[iqbal-lab]

hurrah