o Uݢgr @sXddlmZddlZddlZddlmZmZddlmZm Z m Z m Z ddl Z ddlmmZer?ddlmZGddde ZdZd Zd Zd Zed eZed eZed d Zed d ZedZdZdZGddde Z dGddZ!dGddZ"dHddZ#dHddZ$dHd d!Z%dId%d&Z&dJd+d,Z'dKd1d2Z(dLd6d7Z)d8d9Z*d:d;Z+dMdEdFZ,dS)N) annotationsN)IterableSequence) TYPE_CHECKINGClassVar NamedTupleProtocol)tablesc@seZdZUded<dS)NamedTupleProtocolzClassVar[Sequence[str]]_fieldsN__name__ __module__ __qualname____annotations__rrm/oak/stanford/groups/akundaje/marinovg/programs/cellranger-9.0.1/lib/python/cellranger/analysis/clustering.pyr s r Zgene_expressionZantibody_captureZpeaksZkmeans_Z graphclustZ _graphclustZkmedoidsZcelltypec@s>eZdZUded<ded<ded<ded<ded <ded <d S) CLUSTERINGz#np.ndarray[int, np.dtype[np.int64]]clustersznp.int64 num_clustersz np.float64 cluster_scorez np.bytes_clustering_typeglobal_sort_key descriptionNr rrrrr$s  r cluster_typestr cluster_paramintcCs|tksJd|S)Nz_%d)CLUSTER_TYPE_KMEANSrrrrrformat_legacy_clustering_key-s r!cCs~|tkrtdtd|dS|tkrdtt|fS|tkr%dtt|fS|tkr/dt|fS|ttt fvr8|St d|)zJGenerate a machine-readable string that describes a particular clustering.rZ _clustersz%s_%s_%d_clustersz%s_%d_clusterszUnsupported cluster type: ) r GEX_PREFIXCLUSTER_TYPE_ANTIBODY_KMEANS AB_PREFIXCLUSTER_TYPE_ATAC ATAC_PREFIXCLUSTER_TYPE_KMEDOIDSCLUSTER_TYPE_GRAPHCLUSTCLUSTER_TYPE_ATAC_GRAPHCLUSTCLUSTER_TYPE_CELLTYPES ValueErrorr rrrformat_clustering_key2s r,clustering_keycCs(ttj|}ddd|D}t|S)aGiven clustering keys, output the integer number of clusters. Keys are of a form like _antibody_capture_kmeans_10_clusters or _gene_expression_kmeans_3_clusters. >>> _parse_number_of_clusters("_antibody_capture_kmeans_10_clusters") 10 >>> _parse_number_of_clusters("_gene_expression_kmeans_3_clusters") 3 >>> _parse_number_of_clusters("_peaks_kmeans_5_clusters") 5 cSsg|]}|qSrr).0srrr Tsz-_parse_number_of_clusters..)filterrisdigitjoinr)r- n_clustersrrr_parse_number_of_clustersFs r6cCs|tttfvr |dfS|trt|}t|fS|tr%t|}t|fS|t r2t|}t |fS|t rE| d\}}}t t |fS|tkrM|dfStd|)z,Parse the output of format_clustering_key().rrz0Unsupported clustering type for clustering key: )r(r) CLUSTER_TYPE_ANTIBODY_GRAPHCLUST startswithr"r6rr$r#r&r%r'splitrr*r+)r-r5rrrrparse_clustering_keyXs*     r:cCst|\}}|tkr dS|tkrdS|tkrd|S|tkr"d|S|tkr*d|S|tkr2d|S|tkr8dStd|d |) z)Make a cluster_key string human-readable.zGene Expression Graph-basedzPeaks Graph-basedzGene Expression K-means (K=%d)zAntibody Capture K-means (K=%d)zPeaks K-means (K=%d)zK-medoids (K=%d)Z CelltypeszUnsupported clustering type z for clustering key: ) r:r(r)rr#r%r'r*r+)r-rrrrrhumanify_clustering_keyrs$ r;labels&np.ndarray[int, np.dtype[np._IntType]]returncCs$ttt| }d||S)zRelabel clusters so they are sorted by number of members, descending. Args: labels (np.array(int)): 1-based cluster labels )npargsortbincount)r<orderrrrrelabel_by_sizes rDbase_dir Iterable[int]barcodesIterable[bytes]cCsFtj||}tj|ddtj|d}ddg}t||||dS)NT)exist_okz clusters.csvZBarcodeZCluster)ospathr4makedirs analysis_iosave_matrix_csv)rEr-r<rGout_dirZ clusters_fnheaderrrrsave_clustering_csvs rQf tables.Filelegacy_group_name namedtuplec Cs||j|}t|\}}|tkrdSt||}|||} |jD]$} d|d|d| } | |vr;|j| | | dq!tj d| dq!dS)aSoft-link a legacy-structured (CR 1.2) kmeans subgroup (dest) to a new-style (CR 1.3) subgroup (src). The old-style was a group called 'kmeans' with subgroups named _K. The new-style is a group called 'clustering' with subgroups named kmeans_K_clusters, etc. N/z/_)targetz'Skipped soft-link of legacy dataset to z; node doesn't exist ) create_grouprootr:rr!r Zcreate_soft_linksysstderrwrite) rRZnew_group_namerTrUr-grouprrZ legacy_keyZsubgroupfieldrWrrrcreate_legacy_kmeans_nodess     r_ clustering bc_indices Sequence[int]cCs$t|j||j|j|j|j|jdS)znSelect a subset of barcodes from a clustering object. Args: clustering (CLUSTERS namedtuple) rrrrrr)rrrrrrr)r`rarrrsubselect_barcodessrdcCst|jddS)z>Returns a numpy array containing cell-counts for each cluster.r?N)r@rBr)r`rrrget_cluster_sizessrecCst|dddS)NcSs|jS)N)r)xrrrsz"sort_clusterings..)key)sorted)Z clusteringsrrrsort_clusteringssrjr3np.ndarray[int, np.dtype[np.int64]] | Sequence[int]rrfloatr str | bytesrrc Cs>ttj|tjdt|t|t|t|t|dS)z]Create a clustering namedtuple. Use numpy arrays/scalars to ensure h5 compatibility )dtyperc)rr@asarrayint64float64bytes_rcrrrcreate_clusterings rs)rrrr)r-r)r<r=r>r=)rErr-rr<rFrGrH)rRrSrTrrUr )r`rrarb) rrkrrrrlrrmrrlrrm)- __future__rrJrZcollections.abcrrtypingrrrrnumpyr@cellranger.analysis.ioanalysisiorMZcellranger.wrapped_tablesr r r"r$r&rr#r%r(r)r7r'r*rr!r,r6r:r;rDrQr_rdrerjrsrrrrsD