o Rŀg@sdZGdddZdS)zDApproximate calculation of appropriate thresholds for motif finding.c@sVeZdZdZdddZdddZd d Zd d Zd dZddZ dddZ ddZ dS)ScoreDistributionaClass representing approximate score distribution for a given motif. Utilizes a dynamic programming approach to calculate the distribution of scores with a predefined precision. Provides a number of methods for calculating thresholds for motif occurrences. NcCs|dur#td||_td||j|_||j|_||_ntd|j|_td|j|j|_||j|_|||_|j|jd|_ dg|j|_ d|j | |j <dg|j|_ d|j | |j <|durt ||D] \}}||||jqxdSt|jD]c}dg|j}dg|j} |dd|f}|D]B\} } || } td|| |f| }| | } t|jD]$}|||| |j ||7<| ||| |j || 7<qq||_ | |_ qdS)zInitialize the class.N?)min min_scoremax max_scoreintervallengthn_pointsicmeanstep mo_density _index_diff bg_densityziplog_oddspwmmodify backgroundrangeitemspow_add)selfmotif precisionpssmrlomopositionmo_newbg_newletterscorebgdir+I/var/www/html/myenv/lib/python3.10/site-packages/Bio/motifs/thresholds.py__init__sB       "$zScoreDistribution.__init__rcCst||d|j|jS)Ng?)intr)rxyr+r+r,r4szScoreDistribution._index_diffcCstdt|jd||S)Nr)r rr )rr*jr+r+r,r7szScoreDistribution._addc Csdg|j}dg|j}|D]7\}}||}t|jD](} ||| ||j| ||7<||| ||j| ||7<qq||_||_dS)z%Modify motifs and background density.rN)r rrrrrr) rscoresmo_probsbg_probsr$r%kvr)r*r+r+r,r:s   &( zScoreDistribution.modifycCs@|j}d}||kr|d8}||j|7}||ks |j||jS)zVApproximate the log-odds threshold which makes the type I error (false positive rate).rr)r rrr)rfprr*probr+r+r, threshold_fprFszScoreDistribution.threshold_fprcCs>d}d}||kr|d7}||j|7}||ks|j||jS)zWApproximate the log-odds threshold which makes the type II error (false negative rate).rr)rrr)rfnrr*r9r+r+r, threshold_fnrOszScoreDistribution.threshold_fnrrFcCsr|j}d}d}|||kr%|d8}||j|7}||j|8}|||ks |r1|j||j|fS|j||jS)zMApproximate log-odds threshold making FNR equal to FPR times rate_proportion.rrr)r rrrr)rrate_proportion return_rater*r8r<r+r+r,threshold_balancedXs  z$ScoreDistribution.threshold_balancedcCs|jd|j dS)a"Threshold selection mimicking the behaviour of patser (Hertz, Stormo 1999) software. It selects such a threshold that the log(fpr)=-ic(M) note: the actual patser software uses natural logarithms instead of log_2, so the numbers are not directly comparable. r)r8)r:r)rr+r+r,threshold_patserfsz"ScoreDistribution.threshold_patser)NrNN)r)rF) __name__ __module__ __qualname____doc__r-rrrr:r=r@rAr+r+r+r,r s  #  rN)rErr+r+r+r,s