Software

Sampling Minimal Boolean Expressions (minDNF)

The minDNF method samples minimal boolean expressions in DNF.

Interesting Subspace Mining (SCHISM)

SCHISM finds interesting subspace clusters.

\\\

Graph Pattern Sampling (Output Space Sampling)

• Graph Sampling code: for all, support-biased and discriminative graph pattern sampling

• aRulesSequences (R): R package that contains the cSPADE code (Courtesy: Christian Buchta and Michael Hahsler, Vienna University of Economics and Business Administration).

• cSpade (win) code: same as above, but for the Windows platform (Courtesy: Daniel Diaz, University of Paris 1 - Pantheon Sorbonne).

• Utilities (win) code: same as above, but for the Windows platform (Courtesy: Daniel Diaz, University of Paris 1 - Pantheon Sorbonne).

• BLOSOM code: mug find minimal or-clauses, and-clauses, CNF and DNF expressions, xng finds closed CNF expressions and xug find closed DNF expressions.

Graph Mining & Indexing

[#grail]]

Scalable Graph Reachability Indexing (GRAIL)

GRAIL uses random multiple interval labelings, and a variety of optimizations to perform rapid reachability testing in very large graphs (with millions of nodes and edges).

\\\

• HP Innovation Research Program

• VOGUE code

• Mohammed J. Zaki, Christopher D. Carothers and Boleslaw K. Szymanski, VOGUE: A Variable Order Hidden Markov Model with Duration based on Frequent Sequence Mining. ACM Transactions on Knowledge Discovery in Data, 4(1):Article 5. Jan 2010. (PDF)[==] (BibTeX)

(:htoc start=1 end=2 class=htoneline:)

\\

(:htoc start=1 end=2 class=htoneline:)

(:htoc start=1 end=2:)

Microarray Gene Expression Clustering (TriCluster and MicroCluster)

Biological Sequence Analysis

This section contains code for sequence modeling via Hidden Markov Models, code for structured motif extraction and search, and genome-scale disk-based suffix tree indexing.

Hidden Markov Models: Variable Order HMM with Duration (VOGUE)

Genome Scale Indexing: Disk-based Suffix Trees (Trellis and Trellis+)

Structured Sequence Motifs: Search and Extraction (sMotif and exMotif)

Protein Docking and Partial Shape Matching (ContextShapes)

ContextShapes does rigid-body protein docking. It uses a novel contextshapes data structure to represent local surface regions/shapes on the protein. All critical points on both the receptor and ligand are represented via context shapes, and the best docking is found via pair-wise matching.

complex and informative patterns types such as: Itemsets, Sequences, Trees and Graphs.

(:toc-back:)

(:toc-back:)

(:toc-back:)

(:toc-back:)

(:toc-back:)

Representative Orthogonal Graph Mining (Origami)

Graph Pattern Sampling (MUSK)

(:toc-back:)

This section contains code for mining categorical subspace clusters, shape based clusters, for clustering based on a lower bound on similarity, and a new outlier-based approach for initial cluster seed selection.

Categorical Subspace Clustering (CLICKS)

Arbitrary Shape Clustering (Sparcl)

K-means Initialization (Robin)

(:toc-back:)

Triclusters and Biclusters (TriCluster and MicroCluster)

(:toc-back:)

(:toc-back:)

Disk-based Suffix Trees (Trellis and Trellis+)

(:toc-back:)

(:toc-back:)

(:toc-back:)

• IBM Datagen program: contains the IBM synthetic dataset generator for itemset patterns

(:toc-back:)

VOGUE is a variable order and gapped HMM with with duration. It uses sequence mining to extract frequent patterns in the data. It then uses these patterns to build a variable order HMM with explicit duration on the gap states, for sequence modeling and classification. VOGUE was applied to model protein sequences, as well as a number of other sequence datasets including weblogs.

Genome Scale Indexing

Structured Sequence Motifs: Search and Extraction

sMotif and exMotif

sMotif and exMotif are two complementary for searching and extracting/mining structured sequence motifs DNA sequences. A structured motif consists of simple motifs separated by different gap lengths. The simple motif may be a simple pattern or a position weighted matrix or profile. Given a template structured motif (pattern or profile), sMotif finds all matches in a given set of sequences. On the other hand, exMotif mines novel motifs matching some minimal conditions on the gaps and frequency.

• exMotif code: for structured motif extraction
• sMotif code: for structured motif search

• Yongqiang Zhang and Mohammed J. Zaki, EXMOTIF: efficient structured motif extraction. Algorithms for molecular biology, 1(21). Nov 2006. ((URL)) (PDF) (BibTeX)
• Yongqiang Zhang and Mohammed J. Zaki, SMOTIF: efficient structured pattern and profile motif search. Algorithms for molecular biology, 1(22). Nov 2006. ((URL)) (PDF) (BibTeX)

Protein Structure

Flexible and Non-sequential Protein Structure Alignment (SNAP/STSA and FlexSnap)

SNAP finds non-sequential 3D protein structure alignments. It was initially called STSA (2008-snap). FlexSnap allows the ability to find both flexible and non-sequential allignments.

• SNAP executables
• FlexSNAP (coming soon)

• Saeed Salem, Mohammed J. Zaki and Chris Bystroff, Iterative Non-Sequential Protein Structural Alignment. Journal of Bioinformatics and Computational Biology, 7(3):571-596. Jun 2009. (PDF) (BibTeX)
• Saeed Salem, Mohammed J. Zaki and Chris Bystroff, FlexSnap: Flexible Non-Sequential Protein Structure Alignment. In 9th Workshop on Algorithms in Bioinformatics. Sep 2009. (PDF) (BibTeX)

Protein Docking (ContextShapes)

• ContextShapes code

• Zujun Shentu, Mohammad Al Hasan, Chris Bystroff and Mohammad J. Zaki, Context Shapes: Efficient Complementary Shape Matching for Protein-Protein Docking. Proteins: Structure, Function and Bioinformatics, 70(3):1056-1073. Feb 2008. (PDF)[==] (BibTeX)

Protein Indexing (PSIST)

PSIST uses suffix trees to index protein 3D structure. It first converts the 3D structure into a structure-feature sequence over a new structural alphabet, which is then used to index protein structures. The PSIST index makes it very fast to query for a matching structural fragment.

• PSIST code

• Relevant Publications
  • Feng Gao and Mohammed J. Zaki, PSIST: A Scalable Approach to Indexing Protein Structures using Suffix Trees. Journal of Parallel and Distributed Computing, 68(1):55-63. Jan 2008. (PDF) (BibTeX)

Protein Folding Pathways (UNFOLD)

UNFOLD uses a recursive min-cut on a weighted secondary structure element graph to predict the sequence of protein (un)folding events.

Real and Synthetic Datasets

The section contains various synthetic and real datasets used in some of the papers related to itemset, sequence, tree and XML mining.

Frequent Boolean Expressions (BLOSOM)

The BLOSOM framework allows one to mine arbitrary frequent boolean expressions include AND clauses (itemsets), OR clauses, and CNF/DNF expressions. It focuses on mining the minimal boolean expressions.

• BLOSOM code: mug find minimal or-clauses, xng finds minimal CNF expressions and xug find minimal DNF expressions.
• Charm-L: this code can mine all minimal and closed and-clauses, i.e., all minimal and closed frequent itemsets.

• Relevant Publications
  • Lizhuang Zhao, Mohammed J. Zaki and Naren Ramakrishnan, BLOSOM: A Framework for Mining Arbitrary Boolean Expressions. In 12th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. Aug 2006. (PDF) (BibTeX)

Itemset and Sequence Utilities

Provides the utilities needed for Eclat, Charm/Charm-L, GenMax, Spade and cSpade.

Tree Mining

Real and Synthetic Datasets

The section contains various synthetic and real datasets used in some of the papers related to itemset, sequence, tree and XML mining.

• [[Path:/~zaki/software/IBM-datagen.tgz | IBM Datagen program]: contains the IBM synthetic dataset generator for itemset patterns.
• Tree Generator: contains the synthetic tree generator described in (2005-treeminer:tkde).
• Real Datasets: contains various real itemset datasets like chess, connect, mushroom, pumsb and so on, used in the papers on frequent, closed and maximal itemset mining.
• CSLOGS data: The CSLOGS data was used for (2005-treeminer:tkde).
• Xrules Log Data: The log data used for XML classification in (2006-xrules:mlj).

Itemset and Sequence Utilities

• Utilities code: provides exttpose, getconf, and makebin

Frequent Boolean Expressions (BLOSOM)

The BLOSOM framework allows one to mine arbitrary frequent boolean expressions include AND clauses (itemsets), OR clauses, and CNF/DNF expressions. It focuses on mining the minimal boolean expressions.

Graph Mining

Origami

Graph Pattern Sampling

TriCluster and MicroCluster

Tricluster is the first tri-clustering algorithm for microarray expression clustering. It builds upon the new microCluster bi-clustering approach. Tricluster first mines all the bi-clusters across the gene-sample slices, and then it extends these into tri-clusters across time or space (depending on the third dimension). It can find both scaling and shifting patterns.

• Tricluster code
• microCluster code

• Lizhuang Zhao and Mohammed J. Zaki, TriCluster: An Effective Algorithm for Mining Coherent Clusters in 3D Microarray Data. In ACM SIGMOD Conference on Management of Data. Jun 2005. (PDF) (BibTeX)
• Lizhuang Zhao and Mohammed J. Zaki, MicroCluster: An Efficient Deterministic Biclustering Algorithm for Microarray Data. IEEE Intelligent Systems, 20(6):40-49. Nov/Dec 2005. (PDF) (BibTeX)

Hidden Markov Models (HMM)

Variable Order HMM with Duration (VOGUE)

VOGUE is a variable order and gapped HMM with with duration. It uses sequence mining to extract frequent patterns in the data. It then uses these patterns to build a variable order HMM with explicit duration on the gap states, for sequence modeling and classification.

• VOGUE code

• Invalid BibTex Entry!

Genome Scale Indexing

Trellis and Trellis+

Trellis is a disk-based suffix tree indexing methods (with suffix links) that is capable of indexing the entire human genome on a commodity PC with limited memory. Trellis+ extends Trellis by further removing some memory limitations by using a novel guide suffix tree in memory.

• Trellis code: input sequence must be in memory
• Trellis+ code: removes all memory limitations

• Benjarath Phoophakdee and Mohammed J. Zaki, Genome-scale Disk-based Suffix Tree Indexing. In ACM SIGMOD International Conference on Management of Data. Jun 2007. (PDF) (BibTeX)
• Benjarath Phoophakdee and Mohammed J. Zaki, TRELLIS+: An Effective Approach for Indexing Genome-Scale Sequences using Suffix Trees. In 13th Pacific Symposium on Biocomputing. Jan 2008. (PDF) (BibTeX)

Protein Structure

Flexible and Non-sequential Protein Structure Alignment (SNAP/STSA and FlexSnap)

SNAP finds non-sequential 3D protein structure alignments. It was initially called STSA (2008-snap). FlexSnap allows the ability to find both flexible and non-sequential allignments.

• SNAP executables
• FlexSNAP (coming soon)

• Saeed Salem, Mohammed J. Zaki and Chris Bystroff, Iterative Non-Sequential Protein Structural Alignment. Journal of Bioinformatics and Computational Biology, 7(3):571-596. Jun 2009. (PDF) (BibTeX)
• Saeed Salem, Mohammed J. Zaki and Chris Bystroff, FlexSnap: Flexible Non-Sequential Protein Structure Alignment. In 9th Workshop on Algorithms in Bioinformatics. Sep 2009. (PDF) (BibTeX)

Protein Docking (ContextShapes)

• ContextShapes code

• Zujun Shentu, Mohammad Al Hasan, Chris Bystroff and Mohammad J. Zaki, Context Shapes: Efficient Complementary Shape Matching for Protein-Protein Docking. Proteins: Structure, Function and Bioinformatics, 70(3):1056-1073. Feb 2008. (PDF)[==] (BibTeX)

Protein Indexing (PSIST)

PSIST uses suffix trees to index protein 3D structure. It first converts the 3D structure into a structure-feature sequence over a new structural alphabet, which is then used to index protein structures. The PSIST index makes it very fast to query for a matching structural fragment.

• PSIST code

• Feng Gao and Mohammed J. Zaki, PSIST: A Scalable Approach to Indexing Protein Structures using Suffix Trees. Journal of Parallel and Distributed Computing, 68(1):55-63. Jan 2008. (PDF)[==] (BibTeX)

\\

Frequent Itemsets (Eclat)

Closed Itemsets (Charm, Charm-L) and Non-redundant Rule Generation

Maximal Itemsets (GenMax)

Frequent Boolean Expressions (BLOSOM)

The BLOSOM framework allows one to mine arbitrary frequent boolean expressions include AND clauses (itemsets), OR clauses, and CNF/DNF expressions. It focuses on mining the minimal boolean expressions.

Frequent Sequences (Spade)

Sequence Constraints (cSpade)

SLEUTH

XML Classification (XRules)

XRules uses frequent tree mining to mine discriminative patterns from tree-structured XML documents.

Graph Mining

• MUSK code (coming soon): for maximal patterns
• Graph Sampling code (coming soon): for all, support-biased and discriminative patterns

Clustering

This section contains code for mining categorical subspace clusters, shape based clusters, for clustering based on a lower bound on similarity, and a new outlier-based approach for initial cluster seed selection.

CLICKS

CLICKS finds subspace clusters in categorical data using a k-partite clique mining approach.

• CLICKS code

• Mohammed J. Zaki, Markus Peters, Ira Assent and Thomas Seidl, CLICKS: An Effective Algorithm for Mining Subspace Clusters in Categorical Datasets. Data and Knowledge Engineering, 60(1):51-70. Jan 2007. (PDF)[==] (BibTeX)

Sparcl

Sparcl finds shape-based clusters. It uses a two step approach: in the first step we select a relatively large number of candidate centroids (via ROBIN) to find seed clusters via the K-means algorithm and in the second step we use a novel similarity kernel to merge the initial seed clusters to yield the final arbitrary shaped clusters.

• SPARCL code (coming soon)

• Vineet Chaoji, Mohammad Al Hasan, Saeed Salem and Mohammed J. Zaki, SPARCL: An Effective and Efficient Algorithm for Mining Arbitrary Shape-based Clusters. Knowledge and Information Systems, 21(2):201-229. Nov 2009. (PDF)[==] (BibTeX)

Robin

ROBIN uses a new local outllier factor based initial seed selection to improve k-means style clustering.

• ROBIN code (coming soon)

• Mohammad Al Hasan, Vineet Chaoji, Saeed Salem and Mohammed J. Zaki, Robust Partitional Clustering by Outlier and Density Insensitive Seeding. Pattern Recognition Letters, 30(11):994-1002. Aug 2009. (PDF)[==] (BibTeX)

SLEUTH extends the TreeMiner methodology to mine all frequent embedded or induced as well as ordered or unordered tree patterns.

---

Origami uses random walks over the graph partial order to mine a representative sample of the maximal frequent subgraph patterns. It next selects only the orthogonal representative patterns via a local clique-finding method.

• Origami code (coming soon)

• Vineet Chaoji, Mohammad Al Hasan, Saeed Salem , Jeremy Besson and Mohammed J. Zaki, ORIGAMI: A Novel and Effective Approach for Mining Representative Orthogonal Graph Patterns. Statistical Analysis and Data Mining, 1(2):67-84. Jun 2008. (PDF)[==] (BibTeX)

Graph Pattern Sampling

Whereas Origami can mine a sample of maximal graph patterns it does not provide any uniformity guarantee. MUSK (2009-musk) proposes a Markov Chain Monte Carlo based approach to guarantee a uniform sample of all maximal patterns. The MCMC approach was further extended in (2009-graphsampling) to mine a sample of all frequent patterns, to mine support-biased patterns and also to mine a sample of discriminative patterns.

• MUSK code: for maximal patterns
• Graph Sampling code: for all, support-biased and discriminative patterns

• Mohammad Al Hasan and Mohammed J. Zaki, MUSK: Uniform Sampling of k Maximal Patterns. In 9th SIAM International Conference on Data Mining. Apr 2009. (PDF) (BibTeX)
• Mohammad Al Hasan and Mohammed J. Zaki, Output Space Sampling for Graph Patterns. Proceedings of the VLDB Endowment (35th International Conference on Very Large Data Bases), 2(1):730-741. 2009. (PDF) (BibTeX)

\\

Treeminer uses the vertical scope-list representation to mine frequent sequences (2005-treeminer:tkde). Treeminer counts all embeddings, whereas Treeminer-D counts only distinct occurrences, which can be more appropriate for some datasets.

• TreeMiner code: vtreeminer is the TreeMiner code, whereas htreeminer is the PatternMatcher code as mentioned in the paper below.

SLEUTH extends the TreeMiner methodology to mine all frequent embedded or induced as well as ordered or unordered tree patterns.

• SLEUTH code

• Mohammed J. Zaki, Efficiently Mining Frequent Embedded Unordered Trees. Fundamenta Informaticae, 66(1-2):33-52. Mar/Apr 2005. (PDF)[==] (BibTeX)

• You also need to download utilities like getconf, exttpose from Utilities section.

TreeMiner and TreeMiner-D

Treeminer uses the vertical scope-list representation to mine frequent sequences {[2005-treeminer:tkde]}. Treeminer counts all embeddings, whereas Treeminer-D counts only distinct occurrences, which can be more appropriate for some datasets.

• TreeMiner code
• TreeMiner-D code

• Mohammed J. Zaki, Efficiently Mining Frequent Trees in a Forest: Algorithms and Applications. IEEE Transactions on Knowledge and Data Engineering, 17(8):1021-1035. Aug 2005. (PDF)[==] (BibTeX)

• Charm code
• Charm-L and Non-redundant Rules code

• GenMax code

Spade uses the vertical format for mining the set of all frequent sequences from a dataset of many sequences. It also mines the frequent sequences of itemsets.

• Spade code
• You also need to download utilities like getconf, exttpose from Utilities section.

• Mohammed J. Zaki, SPADE: An Efficient Algorithm for Mining Frequent Sequences. Machine Learning Journal, 42(1/2):31-60. Jan/Feb 2001. (PDF)[==] (BibTeX)]

cSpade mines constrained frequent sequences. The constraints can take the form of length or width limitations on the sequences, minimum or maximum gap constraints on consecutive sequence elements, applying a time window on allowable sequences, incorporating item constraints, and finding sequences predictive of one or more classes. The class specific sequences can be used for sequence classification as described in {[2000-featuremine:is]}.

• cSpade code

• Mohammed J. Zaki, Sequences Mining in Categorical Domains: Incorporating Constraints. In 9th ACM International Conference on Information and Knowledge Management. Nov 2000. (PDF)[==] (BibTeX)

• You also need to download utilities like getconf, exttpose from Utilities section.

• You also need to download utilities like getconf, exttpose from Utilities section.

• You also need to download utilities like getconf, exttpose from Utilities section.

• Eclat code
• You also need to download utilities like getconf, exttpose from Utilities section.

Charm-L adds the ability to construct the entire frequent concept lattice (also called as the iceberg lattice), that is, it adds the links

• You also need to download utilities like getconf, exttpose from Utilities section.

Genmax mines all maximal frequent itemsets via a backtracking approach with progressive focusing.

---

BLOSOM

CLICKS

Sparcl

Robin

Xrules

TriCluster

MicroCluster

VOGUE

Trellis

Snap

FlexSnap

ContextShapes

PSIST

Datasets

DMTL implements all algorithms following the vertical data mining approach. For itemsets, the implementation follows the Eclat approach (without diffsets). For sequences it follows the Spade approach. For trees it implements the SLEUTH framework, which allows one to mine embedded/induced and ordered/unordered trees. Finally, the graph mining framework uses a novel vertical approach.

Itemset Mining and Association Rules

The section contains the code for mining all frequent itemsets, all closed itemsets, and all maximal frequent itemsets. It also includes the code for constructing the concept lattice (or iceberg lattice) and generating nonredundant association rules.

(1997-eclat), combined with the diffsets improvement (2003-diffsets). The code also contains the maxeclat, clique, and maxclique approaches mentioned in

• Path:/~zaki/software/ECLAT.tar.gz

Charm mines all the frequent closed itemsets as described in (2002-charm). Charm-L adds the ability to construct the entire frequent concept lattice, that is, it adds the links between all sub/super-concepts (or closed itemsets) (2005-charm:tkde). This ability is used to mine the non-redundant association rules (2004-nonredundant:dmkd).

• Mohammed J. Zaki and Karam Gouda, Fast Vertical Mining Using Diffsets. In 9th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. Aug 2003. (PDF)[==] (BibTeX)

Charm, Charm-L and Non-redundant Rule Generation

Charm mines all the frequent closed itemsets as described in {[]}.

• Mohammed J. Zaki, Scalable Algorithms for Association Mining. IEEE Transactions on Knowledge and Data Engineering, 12(3):372-390. May/Jun 2000. (PDF)[==] (BibTeX)

• Invalid BibTex Entry!
• Mohammed J. Zaki and Karam Gouda, Fast Vertical Mining Using Diffsets. In 9th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. Aug 2003. (PDF) (BibTeX)

Eclat uses the original vertical tidset approach for mining all frequent itemsets (1997-eclat), combined with the diffsets improvement (2003-diffsets). The code also contains the maxeclat, clique, and maxclique approaches mentioned in (2000-eclat:tkde).

Spade

cSpade

TreeMiner

TreeMiner-D

SLEUTH

DMTL implements all algorithms following the vertical data mining approach. For itemsets, the implementation follows the Eclat approach (without diffsets). For sequences it follows the Spade approach. For trees it implements the SLEUTH framework, which allows one to mine embedded/induced and ordered/unordered trees. Finally, the graph mining framework uses a novel vertical approach.

Itemset Mining

The section contains the code for mining all frequent itemsets, all closed itemsets, and all maximal frequent itemsets.

Eclat

Eclat uses the original vertical tidset approach for mining all frequent itemsets Invalid BibTex Entry!, combined with the diffsets improvement Invalid BibTex Entry!. The code also contains the maxeclat, clique, and maxclique approaches mentioned in Invalid BibTex Entry!.

• Download
  • Download Page
• Relevant Publications
  • Vineet Chaoji, Mohammad Al Hasan, Saeed Salem and Mohammed J. Zaki, An integrated, generic approach to pattern mining: data mining template library. Data Mining and Knowledge Discovery, 17(3):457-495. Dec 2008. (PDF) (BibTeX)

Charm

Charm-L

GenMax

(:*toc:)

complex and informative patterns types such as: Itemsets, Sequences, Trees and Graphs.

• Download
  • Sourceforge Project Page
  • Sourceforge Download Page
• Relevant Publications
  • Vineet Chaoji, Mohammad Al Hasan, Saeed Salem and Mohammed J. Zaki, An integrated, generic approach to pattern mining: data mining template library. Data Mining and Knowledge Discovery, 17(3):457-495. Dec 2008. (PDF) (BibTeX)

(:*toc:)

Acknowledgments

We gratefully acknowledge the funding from the following agencies that made the research possible: National Science Foundation -- Information and Data Management program (IIS-0092978), and Next Generation Software program (EIA-0103708); and Department of Energy, Office of Science (DE-FG02-02ER25538)

Data Mining Template Library (DMTL)

DMTL is an open-source, high-performance, generic data mining toolkit, written in C++. It provides a collection of generic algorithms and data structures for mining increasingly complex and informative patterns types such as: Itemsets, Sequences, Trees and Graphs.

DMTL utilizes a generic data mining approach, where all aspects of mining are controlled via a set of properties. The kind of pattern to be mined, the kind of mining approach to use, and the kind of data types and formats to mine over are all specified as a list of properties. This provides tremendous flexibility to customize the toolkit for various applications.

Download:

• Sourceforge Project Page
• Sourceforge Download Page

Relevant Publications

• Vineet Chaoji, Mohammad Al Hasan, Saeed Salem and Mohammed J. Zaki, An integrated, generic approach to pattern mining: data mining template library. Data Mining and Knowledge Discovery, 17(3):457-495. Dec 2008. (PDF) (BibTeX)

Itemset Mining

(:*toc-float Table of Contents:)

Charm-L

TreeMiner-D

Boolean Expression Mining

BLOSOM

Clustering

CLICKS

XML Classification

Xrules

Microarray Gene Expression Clustering

TriCluster

MicroCluster

PSIST

Utilities

Datasets

(:*toc-float:)

Disclaimer: The software is provided on an as is basis for research purposes. There is no additional support offered, nor are the author(s) or their institutions liable under any circumstances.

List of Software

(:*toc:)

Generic Pattern Mining

Data Mining Template Library

ItemsetMining

Eclat

Charm

GenMax

Sequence Mining

Spade

cSpade

Tree Mining

TreeMiner

SLEUTH

Graph Mining

Origami

Musk

OSS

Clusering

Sparcl

Robin

Hidden Markov Models

VOGUE

Genome Scale Indexing

Trellis

Protein Structure Alignment

Snap

FlexSnap

Protein Docking

ContextShapes

Protein Indexing

PSIST