def calc_pwm_from_simulations(mo, chem_affinity, n_sims=10000):
include_shape = True if mo.encoding_type == 'ONE_HOT_PLUS_SHAPE' else False
# we add 4 bases to the motif length to account for the shape features
seqs = FixedLengthDNASequences(sample_random_seqs(n_sims, 4+mo.motif_len))
affinities = -seqs.score_binding_sites(mo, 'FWD')[:,2]
occs = calc_occ(chem_affinity, affinities)
# normalize to the lowest occupancy sequence
occs /= occs.max()
# give a pseudo count of one to avoid divide by zeros
cnts = np.zeros((4, mo.motif_len), dtype=float)
for seq, occ, aff in izip(seqs, occs, affinities):
for i, base in enumerate(seq.seq[2:-2]):
cnts[base_map[base], i] += occ
# normalize the base columns to sum to 1
return cnts/cnts.sum(0)
def calc_bnd_frac(affinities, chem_pot):
# since the weights default to unfiform, this is the mean on average
return (weights*calc_occ(chem_pot, affinities)).sum()
