consensus calculation is defect/inconsistent

[westram]

Arb has two ways to calculate consensi:

1. NTREE/SAI/Create SAI using …/Consensus
2. Consensus displayed in EDIT4

Code is not shared, because EDIT4 consensus-calculation is mixed with alignment-compression (i.e. hiding gap-columns) in EDIT4 for better performance.
Results of both differ!
Reproduce:

• load attached
• start EDIT4 with config 'consensus' (contains all 41 marked species)
• compare:
  • SAI:consensus (created with method 1.)
  • Consensus of group 'all' (method 2.)
  • ConsAll2 (same; converted to species; use with 'View/Show? only differences to selected')

Consensus-settings used: gaps=off; iupac=on,26%; upper/lower=0/75

Current differences in detail:

Pos	SAI	Group	Col.content	Reason
12	M	Y	1x'Y'	'Y' is correct

Current differences with gaps=on:

Pos	SAI	Group	Col.content	Reason
115	g	s	14x'G'+10x'C'+3x'U'	C(37%)+G(51%)=S; 24/41=58%⇒lowercase
120	a	w	14x'A'+3x'C'+6x'G'+9x'U'	A(43%)+U(28%)=W; 23/41=56%⇒lowercase
153	G	g	23x'G'	23/41=56%⇒lowercase
159	R	r	13x'G'+8x'A'+1x'U'+1x'C'	G(56%)+A(34%)=R; 21/41=51%⇒lowercase

More facts:

• if consensus SAI is calculated with 'gaps=on'
  • the differences listed above disappear (i.e. EDIT4 group consensus seems to use gaps=on implicitely
  • starting with pos=154 several positions are uppercase in SAI and lowercase in EDIT4
• make sure remark of CloTyro2 is ignored in method 1!

---

Differences before [14387]:

Pos	SAI	Group	Col.content	Reason
6	G	g	14x'G'+3x'A'+1x'U'	G(77%) ⇒ 'G'
12	M	n	1x'Y'	both wrong, should be 'Y'
21	M	m	2x'A'+1x'C'	A(66%)+C(33%)=M
32	R	r	18x'A'+12x'G'+6x'C'+3x'U'	A(46%)+G(30%)=R; 46+30=76%⇒uppercase
69	G	g	2x'G'	only 'G' ⇒ should be uppercase
115	S	s	14x'G'+10x'C'+3x'U'	C(37%)+G(51%)=S; 37+51=88%⇒uppercase

Differences before [14319]:

Pos	SAI	Group	Col.content	Reason
12	M	-	1x'Y'
21	M	-	2x'A'+1x'C'	A(66%)+C(33%)=M; gaps are off ⇒ '-' is wrong||
69	G	-	2x'G'	gaps are off ⇒ '-' is wrong!

Differences before [14321]:

Pos	SAI	Group	Col.content	Reason
21	M	a	2x'A'+1x'C'	A(66%)+C(33%)=M
115	S	g	14x'G'+10x'C'+3x'U'	C(37%)+G(51%)=S; 37+51=88%⇒uppercase; gaps are off ⇒ 'g' is wrong!
120	w	a	14x'A'+ 3x'C'+6x'G'+9x'U'	A(43%)+T(28%)=W; 43+28=71%⇒lowercase

---

Amino acids

Consensus calculation for amino acids works does completely different things:

• EDIT4 calculates simplified amino acid code (as currently shown in IUPAC info window; i.e. 'A' for any of "PAGST" etc.)
• NTREE consensus calculation does not seem to handle any amino-IUPAC-codes

Common IUPAC codes for amino acids (both supported in ARB; but probably insufficient):

• B = D | N
• Z = E | Q
• X = <any>

​Some sources also mention J = L | I.

[westram]

demonstrate behavior

[westram]

with [14396] (in branch consensus)

[westram]

Milestone renamed