where is the guide tree estimated?

[GavinHuttley]

Lines of code where dbga is estimating the guide-tree.

[xingjianleng]

In the latest implementation, the DistanceMatrix object from cogent3 was used in de Bruijn MSA. The DistanceMatrix was one input parameter for the alignment() function. The guide tree was constructed at the beginning of alignment.

The current implementation for estimating the DistanceMatrix doesn't have any mathematical/biological meaning, especially how k_estimated is calculated. The implementation is at

DBGA/src/dbga/utils.py

Lines 416 to 437 in 0da8dca

	def distance_matrix_prediction(seq_sc: SequenceCollection) -> Any: # pragma: no cover
	"""generate the estimated distance matrix for input sequences using the predicted pairwise similarity
	Parameters
	----------
	seq_sc : SequenceCollection
	the SequenceCollection containing multiple sequences
	Returns
	-------
	Any
	the estimated distance matrix using the predicted pairwise similarity
	"""
	dist_dict = {}
	for seq_name1, seq_name2 in combinations(seq_sc.names, r=2):
	sub_seqs = seq_sc.take_seqs((seq_name1, seq_name2))
	k_estimated = math.ceil(math.log(min(map(len, sub_seqs.seqs)), 4)) + 2
	distance = 1 - predict_p(sub_seqs, k=k_estimated)
	dist_dict[(seq_name1, seq_name2)] = distance
	dist_dict[(seq_name2, seq_name1)] = distance
	dm = DistanceMatrix(dist_dict)
	return dm