ClusteringResult

Full name SMRUCC.genomics.Model.MotifGraph.ProteinStructure.FamilyCluster.ClusteringResult Assembly SMRUCC.genomics.Model.MotifGraph.ProteinStructure Members 18

full result of the unsupervised protein family clustering pipeline. besides the final family assignment and reference sequences, every intermediate artifact (TF-IDF matrix, SVD embedding, KNN edge list) is preserved so that the run can be inspected and debugged.

for the streaming pipeline the large intermediate products (TF-IDF / SVD / KNN) live on disk under ClusteringResult.workDir; the in-memory fields hold the essentials (names, assignments, families, references, vocabulary) while the big matrices can be streamed back on demand through the reader helpers, so the result object itself never forces the whole database into memory.

01 Syntax

SMRUCC.genomics.Model.MotifGraph.ProteinStructure.FamilyCluster.ClusteringResult

02 Methods

Name	Overloads	Summary
FromDirectory	1	re-assemble a ClusteringResult from a completed streaming run.
StreamTfidf	1	stream the TF-IDF sparse vectors (rowIndex, title, columnIndices, values) back from disk
StreamSvd	1	stream the dense SVD embeddings (rowIndex, embedding) back from disk
StreamKnnEdges	1	stream the KNN edges (u, v, weight) back from disk
ToString	1	serialize the result to a plain JSON object (the TF-IDF / SVD matrices are summarized, not dumped in full)

03 Properties

Name	Overloads	Summary
workDir	1	working directory of the streaming run; holds the on-disk intermediate products (tfidf_vectors.coo, svd_vectors.tsv, knn_edges.tsv, family_assignment.tsv, ...).
sequenceNames	1	titles of the protein sequences, aligned across every step (TF-IDF rows, SVD rows, KNN indices, family assignments)
familyAssignments	1	family id assigned to each sequence (aligned with ClusteringResult.sequenceNames)
referenceSequences	1	reference sequence per family id
families	1	the discovered protein families
tfidfMatrix	1	the TF-IDF matrix (row = sequence, column = kmer vocabulary word)
vocabulary	1	the selected kmer vocabulary (column names of the TF-IDF matrix)
svdVectors	1	the TruncatedSVD embedding: one row per sequence, ClusteringResult.svdDims columns
svdDims	1	the number of SVD dimensions
knnEdges	1	undirected KNN edges: (u, v, weight) with u < v and indices aligned to ClusteringResult.sequenceNames
k	1	the kmer length used for feature extraction
topN	1	number of kmer features kept
familyCount	1	count of distinct protein families

04 Members

method FromDirectory #

FromDirectory(String, String, KmerVocabulary, Dictionary(Of Int32, ProteinFamily))

re-assemble a ClusteringResult from a completed streaming run. the essentials (names, assignments, families, references, vocabulary) are loaded into memory; the very large TF-IDF / SVD / KNN arrays stay on disk and can be streamed back via the reader helpers.

method StreamTfidf #

StreamTfidf

stream the TF-IDF sparse vectors (rowIndex, title, columnIndices, values) back from disk

method StreamSvd #

StreamSvd

stream the dense SVD embeddings (rowIndex, embedding) back from disk

method StreamKnnEdges #

StreamKnnEdges

stream the KNN edges (u, v, weight) back from disk

method ToString #

ToString

serialize the result to a plain JSON object (the TF-IDF / SVD matrices are summarized, not dumped in full)

property workDir #

workDir

working directory of the streaming run; holds the on-disk intermediate products (tfidf_vectors.coo, svd_vectors.tsv, knn_edges.tsv, family_assignment.tsv, ...).

property sequenceNames #

sequenceNames

titles of the protein sequences, aligned across every step (TF-IDF rows, SVD rows, KNN indices, family assignments)

property familyAssignments #

familyAssignments

family id assigned to each sequence (aligned with ClusteringResult.sequenceNames)

property referenceSequences #

referenceSequences

reference sequence per family id

property families #

families

the discovered protein families

property tfidfMatrix #

tfidfMatrix

the TF-IDF matrix (row = sequence, column = kmer vocabulary word)

property vocabulary #

vocabulary

the selected kmer vocabulary (column names of the TF-IDF matrix)

property svdVectors #

svdVectors

the TruncatedSVD embedding: one row per sequence, ClusteringResult.svdDims columns

property svdDims #

svdDims

the number of SVD dimensions

property knnEdges #

knnEdges

undirected KNN edges: (u, v, weight) with u < v and indices aligned to ClusteringResult.sequenceNames

property k #

the kmer length used for feature extraction

property topN #

topN

number of kmer features kept

property familyCount #

familyCount

count of distinct protein families