3Di+AA weights / bitfactors

[Proginski]

Hi,

As far as I understand from the help and the code, a specific weight is used for the construction of the AA, and 3Di substitution matrix (one for each of them).
Especially, the foldmason easy-msa --help mentions:
--bitfactor-aa FLOAT AA matrix bit factor [1.100] --bitfactor-3di FLOAT 3Di matrix bit factor [2.100]
So I understand those bitfactors as numbers that multiply the original 3Di or AA matrix.

Since the Foldseek's paper says "The score linearly combines amino acid and 3Di substitution scores with weights 1.4 and 2.1, respectively.", I guess the bitfactor as changed from 1.4 to 1.1 for AA in the meantime ?

Lastly, I found an occurrence of the 1.4 weight at line 225 of lib/foldseek/src/strucclustutils/structurerescorediagonal.cpp:
float aaFactor = (par.alignmentType == LocalParameters::ALIGNMENT_TYPE_3DI_AA) ? 1.4 : 0.0;, but no equivalent for the 2.1 of 3Di
Is it possible that this is only used for the ungapped aligment part ?

Best,

Paul

[milot-mirdita]

FoldMason's bitfactors are set here:

foldmason/src/commons/FoldmasonParameters.cpp

Lines 85 to 86 in 4bd8b8d

	bitFactorAa = 1.1f;
	bitFactor3Di = 2.1f;

We recently optimized FoldMason's parameters (independently from Foldseek) for MSTA generation.

[Proginski]

Thanks !