hapsburg.hmm_inference

Classes

HMM_Analyze

Analyze Class for HMMs.

Functions

`prep_3x3matrix`(t, n_ref)	Prepares the grouped 3x3 Matrix (3rd State: Everything in OTHER ROH State)
`exponentiate_r`(rates, rec_v)	Calculates exponentiation of the rates matrix with rec_v
`print_memory_usage`()	Print the current Memory Usage in mB

Module Contents

class hapsburg.hmm_inference.HMM_Analyze(folder='./Simulated/Example0/', t_model='model', e_model='haploid', p_model='SardHDF5', post_model='Standard', output=True, save=True, cython=True, manual_load=False, lowmem=False, save_fp=True, start=-np.inf, end=np.inf)

Bases: object

Analyze Class for HMMs. This is the main Class, all specific Inference schemes inherit from it and overwrite functions. Contains objects as field for pre-processing, transition as well as emission probabilities. Contains most Parameters (but some of them like the output folders are decided by pre-processing subclass)

folder = ''

output = True

save = True

save_fp = True

n_ref = 20

sanity_checks = True

ref_states = []

ob_stat = []

pCon = []

lowmem = False

overhang = 0

r_map = []

pos = []

v_path = []

posterior = []

iid = ''

ch = 0

start

end

fwd_bkwd = 0

e_model = 'haploid'

t_model = 'model'

p_model = 'SardHDF5'

post_model = 'Standard'

load_objects(iid='', ch=0, c=0.0): Load all the required Objects in right order

load_secondary_objects(c=0.0): Load all secondary objects (but not the pre-processing one)

load_data(iid='', ch=0): Load the External Data

load_emission_model(c=0.0, pCon=[]): Method to load an Emission Model

load_transition_model(): Load the Transition Model

load_preprocessing_model(conPop=[])

load_postprocessing_model()

prepare_rmap(cm=False, min_gap=1e-10, max_gap=0.05): Return the recombination map [in Morgan] Input: Map Positions [l] Return: Rec. Distance Array [l] cm: Whether input is in centimorgan or morgan min_cap: Minimum Gap between Loci

pre_compute_transition_matrix(t, r_vec, n_ref): Precompute and return the full transition Matrix t full Transition Matrix [k,k]. NO LOG STATE r_vec Map Length of Jumps [l]

calc_posterior(save=True, full=False, in_val=0.0001): Calculate the poserior for each path FULL: Wether to return fwd, bwd as well as tot_ll (Mode for postprocessing) in_val: The Initial Probability to copy from one Ind.

calc_posterior_lowmem(save=True, full=False, in_val=0.0001): same functionality as calc_posterior, but a low-memory implmentation. The emission matrix is not pre-computed, but calculated for each marker on the fly the reference panel is stored in the most compact format possible (aka one bit per allele)

compute_tot_neg_likelihood(c, in_val=0.0001): Calculate the poserior for each path in_val: The Initial Probability to copy from one Ind.

optimze_ll_transition_param(roh_trans_params): Calculate and return the log likelihoods for Transitions Parameters roh_trans_params [m]

optimize_ll_contamination_BFGS(init_c): Find MLE of contamination rate by L-BFGS-B.

optimize_ll_contamANDerr(init_c, init_err)

post_processing(save=True): Do the Postprocessing of ROH Blocks Parameters: See in Postprocessing

mmr_call(windowSize=0.001, save=True): Calculate Maximal Match Rate

hapsburg.hmm_inference.prep_3x3matrix(t, n_ref): Prepares the grouped 3x3 Matrix (3rd State: Everything in OTHER ROH State)

hapsburg.hmm_inference.exponentiate_r(rates, rec_v): Calculates exponentiation of the rates matrix with rec_v rates: 2D Matrix. rec_v: Array of length l

hapsburg.hmm_inference.print_memory_usage(): Print the current Memory Usage in mB