DUf_ddlZddlZddlZddlZddlmZddlZ ddZ ddZ ddgfd Z dd Z d fd Z d Z dZddZdZdZdZdZdZdgfdZdS)N) eigdecompAnyAny_vs_MixedTFc t|tjr(t|jdkrt dn; | rt dtj|}n#t dxYw|||||f}||||||f}| |||}|||dkr| rt dtj t|tj z} tj ||d d } tj | } d }|| | | | <| g}nt d | rt d t||}| rt dt||||| \}}|d |dzdz}tj||ztj|z }|||||fS)a Compute the contrast in M between different sets of pixels specified by modality, modality_params or by I, if given. Returns the contrast in diagonals with offset s=0..len(M), and the weighted average over all s. If 'normalize' is True, contrast is normalized to be in [-1,1]. Parameters ---------- M: 2d array Typically a contact matrix. modality: string, (optional, default: "AnyAny_vs_Mixed") Specifies the contrast modality, potentially together ``modality_params``. See :func:`constrast.indicatormat` for valid modalities. modality_params: list, (optional if modality=='AnyAny_vs_mixed' or None) Parameters required by 'modality'. See :func:`contrast.indicatormat()` for valid modalities and required parameters. Typical format: ``[bin_types, Type]``, where bin_types: 1d array types of the bins, e.g. [0,0,1,1,1,0,...] Note: the numerical values identifying the types are inconsequential they are only used to classify the loci into groups. Type: number one of the values in bin_types: singles out a specific type If None, modality_params = [EV>np.nanmean(EV)], with EV computed here: eigdecomp.cis_eig(M, phasing_track=phasing_track)[1][0] I: indicator matrix, optional Specifies the two sets of pixels from which to compute contrast in M. If given, this overrides ``modality`` and ``modality_params``. If None, computed here as ``contrast.indicatormat(modality, modality_params)``. normalize: boolean, optional If True, contr_diags is normalized as follows contr_norm[s] = (contr[s]-1)/(contr[s]+1) = (set1_avintens[s] - set2_avintens[s]) / (set1_avintens[s] + set2_avintens[s]). The weighted mean, contr, is then computed from contr_norm[s]. ignore_diags: integer, optional Number of diagonals in M to be ignored. exclude_nans_from_paircounts: boolean, optional If True, pixels with NaN in M are not counted towards valid pixels. i0, i1: integers, optional Will use (trimmed by len(M)): M[i0:i1,i0:i1] v[i0:i1], phasing_track[i0:i1] Note: EV is computed from restriced M if modality_params is None phasing_track: 1D numpy array, optional len(phasing_track) must be len(M) to flip EVs only used if modality_params is None Returns ------- contr_diags: 1D numpy array, length M Contrast in diagonals with offsets s=0..len(M) contr: float Weighted mean of contrast_diags. np.nansum(contrast*p)/np.nansum(p) with weights (#set1_pixels(s)*#set2_pixels(s))**.5 I: 2D array The computed (or supplied) indicator matrix. modality_params: list The computed (or supplied) modality_params useful if modality_params was computed here. additional_info: list [ set1_intens: 1D array, nansum(diag(I*M,s)), i.e. total intensity in set1 for all s set1_pixels: 1D array, nansum(diag(I)) i.e. number of valid pixels in set1 for all s set2_intens: 1D array, nansum(diag((1-I)*M)), i.e. total intensity in set2 for all s set2_pixels: 1D array, nansum(1-diag(I)) i.e. number of valid pixels in set2 for all s ] z,M was an array, but dimensionality was not 2z... loading M ...zcmap could not be loadedNrz... getting types from EV...) phasing_trackrrc2|tj|kSNnpnanmeanxs W/var/www/html/software/conda/lib/python3.11/site-packages/cooltools/sandbox/contrast.pyz contrast_diags..sq2:a==/@zmodality_params=None only allowed with modality='AnyAny_vs_Mixed'. For all other modalityes modality_params has to be given (a list). See contrast.indicatormatrix() for valid modalities and required modality_params.z%... constructing indicator matrix ...z... computing contrast ...) ignore_diagsexclude_nans_from_paircounts normalizeg?) isinstancer ndarraylenshape ValueErrorprintjoblibloadonesnanrcis_eigisfinite indicatormatcontrast_diags_indicatormatrixnansum)Mmodalitymodality_paramsIrri0i1rrverbose bin_typesEVval_indsget_bin_identities contr_diagsadd_infopcontrs rcontrast_diagsr5s8V!RZ  9 qw<<1  KLL L  9 +)*** AAA 9788 8 "R%B,A} beRUlO %be,  y  ",,,:8999GCFFOObf4 &q FFFqI!L;r??%@%@"&8&8H&F&F (##,+ 0  ; 9 : : : ? 3 3, *+++:  !%A K  hqk!  A IkAo & &1 5E q/8 ;;s %A**A;c 8d}tj||j|ds'tjd|tj|}|tj|}d|D} tj| rtd| |j |j kstdt|} tj | t} |r!tj| tj|<tj| tjz} tj| tjz} tj| tjz}tj| tjz}|r=t#dtjd | d  t&}t)|| D]}|r9||vr5t#d t'|| z d ztj| |}tj||}tj|| |<tjd |z ||<tj||z| |<tj|d |z z||<tjd5| | z }||z }|dur ||z ||zz }n||z }dddn #1swxYwY| | ||g}||fS)a Computes, for all upper diagonals, the 'contrast' in M, namely the ratio of average intensities (values) in M in two sets of pixels. More formally, for diagonals with offsets s=0..len(M): contr[s] = _set1 / _set2) If normalize is True, the contrast is normalized as follows: contr_norm[s] = (conts[s]-1)/(contr[s]+1) = (_set1 - _set2) / (_set1 + _set2) The normalized contrast is between -1 and 1. The two sets of pixels are specified by the indicator matrix I as follows: set1: all pixels (i,j) with I[i,j]=1 set2: all pixels (i,j) with I[i,j]=0 neither: all pixels (i,j) with I[i,j]=nan If exclude_nans_from_paircounts==True, pixels M(i,j)=nan entail I(i,j)=nan (such that those pixels don't count towards neither set1 nor set2). Parameters ---------- M: 2D numpy array the data, eg. a Hi-C contact matrix I: 2D numpy array indicator matrix specifying the two sets of pixels can contain only - zeros (for set1), - ones (for set2), - NANs (neither) otherwise a value error is raised must have same shape as M ignore_diags: int the entries [0:ignore_diags] will be NANs in all returns exclude_nans_from_paircounts: boolean if True, I(i,j) gets nans for all pixels (i,j) where M is nan normalize: boolean, optional if True, contrast is normalized as follows: contr_norm[s] = (contr[s]-1)/(contr[s]+1) = (set1_avintens[s] - set2_avintens[s]) / (set1_avintens[s] + set2_avintens[s]) Returns ------- contr_diags: 1D array ratio of mean values in M of set1 and set2, by diagonal s additional_info: list [ set1_intens: 1D array, nansum(diag(I*M,s)), i.e. total intensity in set1 for all s set1_pixels: 1D array, nansum(diag(I)) i.e. number of valid pixels in set1 for all s set2_intens: 1D array, nansum(diag((1-I)*M)), i.e. total intensity in set2 for all s set2_pixels: 1D array, nansum(1-diag(I)) i.e. number of valid pixels in set2 for all s ] Notes ----- * Lower halves of M and I are ignored but a warning is thrown if M is not symmetric. * The bare contrast can be computed from the normalized one as contr = (contr_norm + 1)/(contr_norm - 1). Examples -------- M = [3 2 1 I = [1 1 0 . 1 1 . 1 0 . . 1] . . nan] The resulting below quantities are by diagonal with increasing offset s=0..len(M): set1_intens = [4, 2, 0] set1_pixels = [2, 1, 0] => set1_avintens = [2, 2, nan] set2_intens = [0, 1, 1] set1_pixels = [0, 1, 1] => set2_avintens = [nan, 1, 1] => contrast_diags_indicatormatrix(M, I, ignore_diats=0) = set1_avintens / set2_avintens = [nan, 2, nan] (indeed, only 1st off-diagonal has both types, others have to be nans) contrast_diags_indicatormatrix(M, I, ignore_diats=0, normalize=True) = (set1_avintens - set2_avintens) / (set1_avintens + set2_avintens) = [nan, 1/3, nan] indeed, only 1st off-diagonal has both types, others have to be nans g:0yE>T)atol equal_nanz9M is not symmetric to within {}, I'm using the upper halfcg|]}|dv| S))gg?).0is r z2contrast_diags_indicatormatrix..1s";;;q ':':1':':':rzKI is not an indicator matrix: it contains value(s) other than {0,1,np.nan}:z M and I must have the same shapez... ... starting to loop over sr z... ... {}% donedrignore)invalidN)r allcloseTwarningswarnformatuniqueisnananyrrrcopyastypefloatr zerosrlinspaceintrangediagr%errstate)r&r)rrrr,tolIunanIu illegalvalsLIuse set1_pixels set1_intens set2_pixels set2_intensreport_progres_pointssdiagIdiagM set1_avintens set2_avintensr4additional_infos rr$r$s;b C ;q!#C4 8 8 8  J Q QRU V V   1B  } E;;e;;;K vk  <     7ag  ;<<< AA 71::  U # #D$#FRXa[[(1++&K(1++&K(1++&K(1++&KB /000 " Aq" 5 5 < > >a  1 5)) A1u9-- A55=11 A5AI#677 A X & & & 2 2#k1 #k1   "]2}}7TUEE!M1E 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2#KkJO / !!s LL L cDt|}g}tt|D]q}tj|}t|D]6}tjtj||||k||<7||r|S)a count occurrances of each v in vals in all diagonals of I with offsets s=0..len(I) Returns ------- list of 1D arrays of floats, one for each v in vals, each of len=len(I) default: count 1s and 0s in each diagonal: return [S1,S0] with: S1: nansum(diag(I,s)==1) for each offset s=0..len(I) S0: nansum(diag(I,s)==0) for each offset s=0..len(I) )rrPr rMr%rQappend)r)valsrW diagcountsr<Sr^s rrgrgps AAJ 3t99   HQKKq 7 7A9RWQ]]d1g566AaDD! rcRtj|}tj|tj|}atjtjtjtjfd|D}n|}|S)a Find unique and finite values (bin types) in v. If vals is not None, restrict to types in vals (floats and ints are considered identical). Parameters ---------- v: sequence of numbers a vector of "types" used to construct indicator matrices vals: sequence of numbers, optional default: unique and finite values in v Returns ------- valid_types: 1d array NcDg|]}|vtj||Sr:)r r")r;trfs rr=zget_types..s+!S!S!Sa4iiBKPQNNi!iiir)r asarrayrGr")vrftypes valid_typess ` r get_typesrps& 1 A Ia A' ( (E z$ybk$//011j!S!S!S!SU!S!S!STT  rc2|tj|kSr r rs rrrsQA5Frctjt|tjz}tj|}|||||<|S)z discretizes a (quasi)continuous track (eg. an EV) using the supplied lambda function by default: split by nanmean returns: discretized track, non-finite values are left intact )r rrr r")rmr0v_discr/s rdiscretize_trackrtsIWSVV__rv %F{1~~H))!H+66F8 Mrc@tj|}|dz |dzz SNrr rlrms rrr" 1 A Ea!e rc@tj|}d|zd|z z Srvrwrxs r normalize_invr{ryrc.tttttt d}d}|D] }||zdz} |td|zdS||vrt|dz|z|||}|S)a get indicatormatrix I with a given modality by calling the function specified by 'modality' Parameters ---------- modality: string, optional valid modalities are listed below in dict 'valid_modalities' if modality is None: print a list of valid modalities and return nothing params: list, optional have to match the parameters of the called indicator matrix functions Returns ------- I: 2d array the computed indicator matrix )rTypeType_vs_MixedTypeType_vs_TypeOtherTypeType_vs_NontypeNontypeTypeType_vs_RestSegments_vs_Rest Nz,valid modalities are: ===================== z1 is not a valid modality. Valid modalities are: ) indicatormat_AnyAny_vs_Mixedindicatormat_TypeType_vs_Mixed"indicatormat_TypeType_vs_TypeOther'indicatormat_TypeType_vs_NontypeNontypeindicatormat_TypeType_vs_Restindicatormat_Segments_vs_Restkeysrr)r'paramsvalid_modalitiesvalid_modalities_nameskr)s rr#r#s,8;!C&M99    " " $ $CC!7!!;d!B >AWWXXX',,.... B C$ %   #"F+A Hrctj|}|j}t|}tj|tjz}d|tj|<tj||}|D]J}tj|tjz}d|||k<tj||}d||dk<K|S)a Parameters ---------- v: sequence of numbers a vector of "types" used to construct indicator matrices types are the unique and finite values in v Returns ------- I: 2d array indicator matrix with the following properties I[i,j]=1 if v[i]==v[j], i.e. thistype-to-thistype for any type I[i,j]=nan if v[i]==nan or v[j]==nan, i.e. excluding invalid bins I[i,j]=0 otherwise, i.e. thistype-to-othertype for valid types rr)r rlsizerprMr rHouter)rmrWrnv_auxr)thistype I_thistypes rrrs" 1 A A aLLE HQKK"& EE28A;;, A bf$ a8mX 5   !O   HrcBtj|}|j}tj||kst jdt |}tj|tjz}d|||k<tj ||}d||dk<|S)a Parameters ---------- v: sequence of numbers a vector of "types" used to construct indicator matrices Type: number type for which to compute indicator matrix with properties Returns ------- I: 2d array: indicator matrix with the following properties I[i,j]=1 if v[i]==v[j]==val, i.e. thistype-to-thistype I[i,j]=0 if v[i]~=v[j]==val, i.e. sometype-to-othertype I[i,j]=nan otherwise i.e. othertype-to-othertype and invalid bins Type not found in vrr) r rlrrIrDrErrMr rrmTyperWI_AnyAny_vs_Mixedrr)s rrr)s& 1 A A 6!t)  - +,,,4  HQKK"& EE!t) A !A1 Hrctj|}|j}tj||kst jd tj|tjz}d|tj |<tj ||}tj |}d|||k<tj ||} tj|tjz}d|||k<tj ||} tj||ftjz}d|| dk<d|| dk<tj|tj |<|S)a) Parameters ----------- v: sequence of numbers a vector of "types" used to construct indicator matrices Type: number type for which to compute indicator matrix with properties Returns ------- I: 2d array: indicator matrix with the following properties I[i,j]=1 if v[i]==v[j]==val, i.e. thistype-to-thistype I[i,j]=0 if v[i]==val xor v[j]==val i.e. thistype-to-othertype I[i,j]=nan otherwise i.e. othertype-to-othertype and invalid bins rrr)r rlrrIrDrErrMr rndindexr"rHr) rmrrWrrr)r<j I_invalid I_striped I_TypeTypes rrrNs6& 1 A A 6!t)  - +,,, ( bf$rx{{lHUE**  a4iHUE**  bf$a4iXeU++ HaV  rv %)q.*/!#"(9   Hrctj|}|j}tj||kst jdt |d}tj|tjz}d|||k<tj ||}tj|tjz}d|||ktj |z<tj ||}d||dk<|S)a* Parameters ----------- v: sequence of numbers a vector of "types" used to construct indicator matrices Type: number type for which to compute indicator matrix with properties Returns ------- I: 2d array: indicator matrix with the following properties I[i,j]=1 if v[i]==v[j]==val, i.e. thistype-to-thistype I[i,j]=0 if v[i]!=val and v[j]!=val i.e. anything not involving type I[i,j]=nan otherwise i.e. type-to-othertype and invalid bins rrr) r rlrrIrDrErrMr rr")rmrrWrrr)I_auxs rrrs& 1 A A 6!t)  - +,,,5Q77: HQKK"& EE!t) A HQKK"& E*+E19 A &' HUE " "EAeqjM Hrcjtj|}|j}tj||kst jdt |d}tj|tjz}d|tj |<d|||k<tj ||}|S)a2 Parameters ----------- v: sequence of numbers a vector of "types" used to construct indicator matrices Type: number type for which to compute indicator matrix with properties Returns ------- I: 2d array indicator matrix with the following properties I[i,j]=1 if v[i]==v[j]==val, i.e. thistype-to-thistype I[i,j]=nan if v[i]==nan or v[j]==nan, i.e. excluding invalid bins I[i,j]=0 otherwise, i.e. thistype-to-othertype and anytype-to-anyother rrr) r rlrrIrDrErrMr r"rrs rrrs( 1 A A 6!t)  - +,,,5Q77: HQKK"& EE"+a..E!t) A Hrcn||dd}tj|}|gkr!tj|tj|<tj||}g}|D]F\}}t ||}|||f|||||fdz|||||f<Gd||dk<|S)al Parameters ---------- segments: list of tuples [(s0,e0),(s1,e1), ...] where (s,e) are the start and end points of segments (endpoints are exclusive) Note: endpoints are trimmed to L L: integer, optional default: last endpoint bad_bins: list of integers, optional bad bins, rows and comuns get np.nan throughout Returns ------- I: 2d array indicator matrices with the following properties I[i,j]=1 if i and j within same segment I[i,j]=nan if i in bad_bins or j in bad_bins I[i,j]=0 otherwise Nr)r rMr rlrminre)segmentsrWbad_binsrr)segments_trimmedr^es rrrs2 y RLO HQKKE2~~&(fbj""# A&&1 1IIA'''!QqSkAo!A#qs( Aa!eH Hr) rNNrTNNNFF)rTTFr )NN)rDosnumpyr rrrcoolerr5r$rgrprtrr{r#rrrrrrr:rrrs    !%   f<f<f<f