Investigate differences between Semi-supervised and Hybrid

[maximskorik]

hybrid.R is a refactored version of Hybrid method, originally implemented in semi.sup.R. Both scripts follow the same procedure, that being:

1. feature extraction
2. time adjustment
3. feature alignment
4. feature augmenting from a table of known compounds
5. weaker-signal recovery
6. second time adjustment
7. second feature alignment
8. feature augmenting from a table of known compounds

More detailed description on Emory's web.

---

However, the original and refactored tools produce different outputs. hybrid.R tends to detect more features and shift retention times of features that match between two outputs. For features that differ they often do so at just the 3rd or 4th decimal place of m/z. This is apparent by the fact that rounding features to less decimal places leads to more aligned features between the two outputs (example).

---

Since semi.sup.R is a part of apLCMS Two-Step Hybrid, understanding where the two tools differ would make possible to replace semi.sup() with hybrid() and split apLCMS Two-Step Hybrid into separate Galaxy tools.

---

Inputs: https://umsa.cerit-sc.cz/u/hechth/h/unnamed-history-2 QC_1.raw & QC_2.raw

Output difference between two methods:

tool	1st step of two-step method (semi.sup.R)	hybrid
num of features	2770	2964
overlap (mz)	1066	1066
overlap; rounded to 4. place (mz)	1911	1911
overlap; rounded to 3. place (mz)	2690	2690
average rt deviation between matching peaks (s)	36.4

[hechth]

@maximskorik do we have a test comparing semi.sup and hybrid?

[maximskorik]

@maximskorik do we have a test comparing semi.sup and hybrid?

@hechth no. I've never pushed that branch and imprudently deleted it.