3 y^>@sddlmZmZmZmZddlZddlZddlZddlZddl Z ddl Z ddl Z ddl Z ddl Z ddlZddlZddlZddlmZddlZddlZejeZejejGdddZdS))absolute_importdivisionprint_functionunicode_literalsN)GraphCycleExceptionc@sreZdZdZddZddZddZdd Zd d Zd d Z e ddZ e ddZ e ddZ e dddZdS)Gene_splice_modelerz Builds supertranscipts. object instance members: gene_id : str alignments : list of Node_alignment objects cCsP||_t|_tjdj|x,|D]$}|j}tjj|}|jj |q$WdS)a? initialize alignments list with simple single 'alignment' objects with each path as an individual alignment with just its path nodes. params: gene_id : str node_path_obj_list : list of Node_path objects, each Node_path corresponding to an individual Trinity isoform zGene_splice_modeler inputs: {}N) gene_idlist alignmentsloggerdebugformatZget_transcript_nameNode_alignmentZget_single_seq_node_alignmentappend)selfrnode_path_obj_listZ node_path_objtranscript_nameZ alignment_objrW/broad/hptmp/bhaas/trinityrnaseq/Analysis/SuperTranscripts/pylib/Gene_splice_modeler.py__init__&s  zGene_splice_modeler.__init__cCs|jS)N)r)rrrr get_gene_idAszGene_splice_modeler.get_gene_idc Cs8|js,y|jStk r(|jSXn|jSdS)a' method to construct the super transcript. Tries 2 approaches: a. If there isn't an obvious repetitive node structure and so the graph formas a DAG, we build a splice graph and perform topological sorting of the nodes. b. If there is some repetitive structure, we resort to performing a multiple alignment-based method to organize relationships among nodes in isoforms, and the multiple alignment produces the linear ordering for the supertranscript. N)alignment_contains_repeat_nodetopological_order_splice_modelrmultiple_alignment_splice_model)rrrrbuild_splice_modelEs  z&Gene_splice_modeler.build_splice_modelcCsZxT|jD]J}t}x>td|jD],}|j|}|j}||krDdS|j|q"WqWdS)NrTF)r setrangewidthZget_representative_column_node get_loc_idadd)r alignmentloc_idsinode_objloc_idrrrr`s  z2Gene_splice_modeler.alignment_contains_repeat_nodecCs*tjd|j}tj|}x|jD]}tjdt||jd}|jd}tjdt|xtdt |D]}||}|j }|j |||j } tjdt| |dkr||d} |j || j | j } | j | |t |dkrv||d} |j || j | j } | j| qvWq&Wtjdt|tjj|j}tjdt|t}x,tdt |D]}||j }|||<qpWt}t}xj|jD]`}|j|jdd d |D}x.|jdD]}|j }||}|||<qW|j|qWtj|||}tjd t||S) zW Build supertranscript using simpler topological sorting of the nodes. z using topological sort method. z1topological_order_splice_model, input alignment: rz+topological_order_splice_model, node list: zgeneric node: zBefore sorting nodes: zTopologically sorted nodes: cSsg|]}dqS)Nr).0r"rrr szFGene_splice_modeler.topological_order_splice_model..zSplice graph model: )r r rTGraphr strZget_aligned_nodesget_transcript_namesrlenrget_nodeget_seq add_prev_node add_next_nodeTopological_sortZtopologically_sort get_all_nodesdictr rr)rrZgraphr node_listrr"r#r$Z generic_node prev_node_objZprev_generic_node next_node_objZnext_generic_nodeZtopologically_sorted_nodesZaligned_loc_id_posZnew_alignmentsZtranscript_idsZ new_alignmentnodeZnew_idxZsplice_graph_modelrrrrnsR         z2Gene_splice_modeler.topological_order_splice_modelc Cs\tjd|j}t|dkr$|dStj|j}tjdt|xt|dkr2x"tdt|D]}d|||<qbWtt j |}t|}t||}||}||}||} tj || } t } x0tdt|D]}|||fkr| j ||qW| j | | }tjdt|tj|}tjdt|qDWt|dkrTtdjt||dS)z Multiple alignment algorithm for dealing with repeat nodes: For each best matching pair of transcripts (or aligned transcripts), perform alignment, and replace aligned pair with a single alignment object. z using mult alignment method. r%rzSimilarity matrix: z Updated alignments: zHError, should only have one alignment but have {} alignments after merge)r r r r+rcompute_similarity_matrixr)rintnumpyargmaxmerge_alignmentsr r RuntimeErrorr ) rr Zsimilarity_matrixr"Z best_pair_idxnum_alignmentsZbest_pair_idx_1Zbest_pair_idx_2align_aalign_bZ align_mergedZnew_alignment_listrrrrs8        z3Gene_splice_modeler.multiple_alignment_splice_modelc Cst|}tj||fdd}x`td|dD]N}||}x@t|d|D].}||}tjj||}t|}||||<qFWq*W|S)zc similarity matrix indicates number of shared nodes between each pair of isoforms. int_)dtyperr%)r+r:zerosrrZcompute_number_common_nodes) Zalignments_listr>Z sim_matrixr"Zalign_ijZalign_jZ common_nodesZnum_common_nodesrrrr8sz-Gene_splice_modeler.compute_similarity_matrixcCstjdj||t|j}t|j}tj||sFtdj|||j}|j}tjj ||}x&t d|dD]}d||dd<qtWx&t d|dD]}d|d|d<qWxt d|dD]}xt d|dD]}t j ||d||d} ||d|dd| } |||dd} ||d|d} | dkr|| | kr|| | kr|| |||d<d |||d<q| | kr| |||d<d|||d<q| |||d<d|||d<qWqW|} |}| }|}t }x|dks|dkr|||}|j|d}|j|d}t }|d}|d krX|d8}|d8}||}n|dkr|d8}||7}xtt dt|D]}|jd qWnP|dkr|d8}x"t dt|D]}|jd qW||7}ntd j|||j|qW|jtjd t||j|j}t }xFt dt|D]4}t }x|D]}|j||qHW|j|q8Wtjd t|tj|j||}tjdt||S)z Computes a mismatch-free multiple alignment (just matches and gaps) between two Node_alignment objects returns single Node_alignment object containing the contents of aligned align_a and align_b as aligned. zMerging alignments {} and {}zEError, transcripts in alignments to merge are not disjoint: {} and {}r%ZDEL_BrbtZDEL_AscoreZDIAGNzbt: ({},{}), bt_dir not definedzMerged alignment nodes list: zmerged alignment node matrix: zmerged alignment obj: )r r r rr* isdisjointr=r DP_matrixZbuild_DP_matrixrrget_match_scorer Zget_node_LIST_at_column_posr+rreverser)rr)r?r@Ztranscript_names_align_AZtranscript_names_align_BZwidth_aZwidth_bZ dp_matrixr"rDZscore_cell_matchZ score_diagZ score_del_aZ score_del_bmax_iZmax_jZall_merged_alignment_nodes_listZ score_structZ nodes_align_aZ nodes_align_bZ align_nodesZbt_dirxZmerged_transcript_name_listZnode_obj_matrixrowZ node_obj_listZmerged_alignment_objrrrr<s            z$Gene_splice_modeler.merge_alignmentscCsD|j|}|j|}tjj|}tjj|}tj||r;   r) __future__rrrrossysreloggingargparse collectionsr:timer(TNode Node_pathrrr0rH getLoggerrZr addHandler NullHandlerrrrrrs