def ml_encl_FRR_span(in_pref, exp_dir, weights):
arg_dict = load_profile(exp_dir)
in_span = in_pref + '_srp.sam'
in_encl = in_pref + '_er.sam'
in_frep = in_pref + '_frr.sam'
if 'diploid' in arg_dict and arg_dict['diploid'] == 'True':
fn = lambda x: (-weights['frr'] * FRR_sam_likelihood (x, in_frep, arg_dict, weights) + \
-weights['srp'] * span_sam_likelihood(x, in_span, arg_dict, weights) + \
-weights['er' ] * encl_sam_likelihood(x, in_encl, arg_dict, weights))
# for i in [1, 3, 5, 10, 20, 40, 80]:
# print i, fn([i]), encl_sam_likelihood(i, in_encl, arg_dict)
res = minimize(fn, x0 = [10,30], \
method = 'L-BFGS-B', \
bounds = [(1, 280), (1, 280)])
return [int(round(j)) for j in res.x]
else:
fn = lambda x: (-weights['frr'] * FRR_sam_likelihood(x[0], in_frep, arg_dict, weights) + \
-weights['srp'] * span_sam_likelihood(x[0], in_span, arg_dict, weights) + \
-weights['er' ] * encl_sam_likelihood(x[0], in_encl, arg_dict, weights))
# for i in [1, 3, 5, 10, 20, 40, 80]:
# print i, fn([i]), encl_sam_likelihood(i, in_encl, arg_dict)
res = minimize(fn, x0 = [70], \
method = 'L-BFGS-B', \
bounds = [(1, 280)])
return [int(round(j)) for j in res.x]
def ml_enclosing_spanning(in_pref, exp_dir):
arg_dict = load_profile(exp_dir)
in_span = in_pref + '_srp.sam'
in_encl = in_pref + '_er.sam'
in_frep = in_pref + '_frr.sam'
fn = lambda x: (-1 * FRR_sam_likelihood(x[0], in_frep, arg_dict) + \
-1 * span_sam_likelihood(x[0], in_span, arg_dict))
fn_span = lambda x: (-1 * span_sam_likelihood(x[0], in_span, arg_dict))
fn_frep = lambda x: (-1 * FRR_sam_likelihood(x[0], in_frep, arg_dict))
# def diminisher(x, encl_allele):
# val = fn_span_single(encl_allele)
# sig = 1
# amplitude_damper = 0.01
# return amplitude_damper * val * np.exp(-0.5 * ((x - encl_allele[0]) / sig) ** 2)
# fn_span_single_diminished = lambda x: fn_span_single(x) + diminisher(x[0], encl_allele)
# Now try using spanning read pairs:
res = minimize(fn_span, x0 = [30], \
method = 'L-BFGS-B', \
bounds = [(1, 250)])
return [int(round(j)) for j in res.x]
def extract_pre_flank (exp_dir, length): arg_dict = load_profile(exp_dir) exp_name = arg_dict['exp_name'] temp_fa_dir = exp_dir + '/temp/' + exp_name with open(temp_fa_dir + '_prefix.fa', 'r') as f: f.readline() pref = f.readline()[-length - 1:].strip() return pref
def extract_post_flank (exp_dir, length): arg_dict = load_profile(exp_dir) exp_name = arg_dict['exp_name'] temp_fa_dir = exp_dir + '/temp/' + exp_name with open(temp_fa_dir + '_suffix.fa', 'r') as f: f.readline() post = f.readline()[:length].strip() return post
def __init__(self, state, in_pref, weights, exp_dir): self.random_range = 5 self.in_span = in_pref + '_srp.sam' self.in_encl = in_pref + '_er.sam' self.in_frep = in_pref + '_frr.sam' self.weights = weights self.arg_dict = load_profile(exp_dir) super(OptimalAlleleLengthProblem, self).__init__(state) # important!
def likelihood(in_pref, exp_dir, weights, alleles): arg_dict = load_profile(exp_dir) in_span = in_pref + '_srp.sam' in_encl = in_pref + '_er.sam' in_frep = in_pref + '_frr.sam' fn = lambda x: -1 * (-weights['frr'] * FRR_sam_likelihood (x, in_frep, arg_dict, weights) + \ -weights['srp'] * span_sam_likelihood(x, in_span, arg_dict, weights) + \ -weights['er' ] * encl_sam_likelihood(x, in_encl, arg_dict, weights)) return fn(alleles)
def run_sim_annealing(exp_dir, in_pref, weights, oalp_params): arg_dict = load_profile(exp_dir) oalp = OptimalAlleleLengthProblem(oalp_params['init_state'], in_pref, weights, exp_dir) oalp.steps = oalp_params['steps'] oalp.updates = oalp_params['updates'] oalp.Tmax = oalp_params['Tmax'] oalp.Tmin = oalp_params['Tmin'] oalp.copy_strategy = oalp_params['copy_strategy'] state, e = oalp.anneal() return state, e
def ml_FRR_spanning(in_pref, exp_dir, weights):
arg_dict = load_profile(exp_dir)
in_span = in_pref + '_srp.sam'
in_encl = in_pref + '_er.sam'
in_frep = in_pref + '_frr.sam'
nc = 80
fn = lambda x: (-weights['frr'] * FRR_sam_likelihood(x[0], in_frep, arg_dict) + \
-weights['srp'] * span_sam_likelihood(x[0], in_span, arg_dict))
res = minimize(fn, x0 = [70], \
method = 'L-BFGS-B', \
bounds = [(1, 250)])
return [int(round(j)) for j in res.x]
def run_max_likelihood(in_pref, exp_dir, weights, ml_params): arg_dict = load_profile(exp_dir) in_span = in_pref + '_srp.sam' in_encl = in_pref + '_er.sam' in_frep = in_pref + '_frr.sam' if ml_params['dimensions'] == 1: fixed_allele = ml_params['fixed_allele'] fn = lambda x: (-weights['frr'] * FRR_sam_likelihood ([fixed_allele, x[0]], in_frep, arg_dict, weights) + \ -weights['srp'] * span_sam_likelihood([fixed_allele, x[0]], in_span, arg_dict, weights) + \ -weights['er' ] * encl_sam_likelihood([fixed_allele, x[0]], in_encl, arg_dict, weights)) elif ml_params['dimensions'] == 2: fn = lambda x: (-weights['frr'] * FRR_sam_likelihood (x, in_frep, arg_dict, weights) + \ -weights['srp'] * span_sam_likelihood(x, in_span, arg_dict, weights) + \ -weights['er' ] * encl_sam_likelihood(x, in_encl, arg_dict, weights)) res = minimize(fn, x0 = ml_params['init_state'], \ method = 'L-BFGS-B', \ bounds = ml_params['bounds']) return [int(round(j)) for j in res.x]
res = minimize(fn, x0 = [70], \
method = 'L-BFGS-B', \
bounds = [(1, 280)])
return [int(round(j)) for j in res.x]
weights = { 'frr': 0.8, \
'srp': 1.0, \
'er': 1.0, \
'allele_1': 0.5,\
'allele_2': 0.5}
exp_dir = '/storage/nmmsv/expansion-experiments/ATXN3_41_cov60_dist500_DIP/'
# exp_dir = '/storage/nmmsv/expansion-experiments/ATXN7_24_cov50_dist700_hap/'
in_pref_pref = exp_dir + 'aligned_read/nc_'
arg_dict = load_profile(exp_dir)
print exp_dir
in_pref = in_pref_pref + '10'
in_span = in_pref + '_srp.sam'
in_encl = in_pref + '_er.sam'
in_frep = in_pref + '_frr.sam'
# for i in [5, 10, 20, 30, 60, 70]:
# fn = lambda x: (-weights['frr'] * FRR_sam_likelihood (x, in_frep, arg_dict, weights) + \
# -weights['srp'] * span_sam_likelihood(x, in_span, arg_dict, weights) + \
# -weights['er' ] * encl_sam_likelihood(x, in_encl, arg_dict, weights))
# print fn([20, i])
fn = lambda x: (-weights['frr'] * FRR_sam_likelihood (x, in_frep, arg_dict, weights) + \
-weights['srp'] * span_sam_likelihood(x, in_span, arg_dict, weights) + \
-weights['er' ] * encl_sam_likelihood(x, in_encl, arg_dict, weights))
