RNA, API and sample comparison prior

Version 1.4 release

Major technical changes

• Improvements to RNA re-squiggle
  • RNA stall detection and masking/collapsing
    • Only affects events during dynamic programming (raw signal is never removed or ignored)
  • Improved signal normalization
    • Event-based
    • Adapter trimming
  • Updated RNA parameter tuning
    • Better detection of reads leaving reasonable signal to sequence matching path
• Added canonical model prior for sample comparison modified base detection
  • This has improved modified base detection from an AUC of 0.91 and 0.84, for 5mC and 6mA respectively, to 0.99 and 0.94 on a relatively low coverage bacterial sample (~20X).

Official Tombo python API release (documentation here)

• API includes 3 modules: tombo_helper, tombo_stats and resquiggle
• Provides access to base mean and raw signal data from individual reads or across a genomic range
• Provides access to per-read modified base detection statistics
• Key objects include: readData, intervalData, TomboReads, TomboStats and PerReadStats

This release also includes various bug fixes including a change to the statistics file allowing processing of arbitrarily large genomes.

Parameter tuning, specifically the read start identification window size and adaptive bandwidth, should result in improved computational performance for the majority of runs (low quality data or poor reference quality may cause longer run times).