REDUCE Suite

The REDUCE Suite distribution also includes the following auxiliary programs. Simple type the corresponding program name with -h (e.g., Convert2PSAM -h) should provide sufficient information to get one started.

Convert2PSAM

---

As its name suggests, Convert2PSAM is a utility program that converts other commonly used motif (pattern) representations in nucleic acid sequences to PSAM, which is unique to the REDUCE Suite. It can also be used to standardize the various formats to a simplified PWM format for easy communication.

Topo2Dictfile

---

The default topological pattern mechanism can be used to specify sequence motifs in a compact, convenient, and flexible way. However, it defines the motifs implicitly, has length limit (15 non-gap positions), and does not take into consideration of the IUPAC degenerate symbols. As an example, X6 stands for exactly 4^6 = 4096 combinations, from AAAAAA, AAAAAC, ... TTTTTT. Sometimes, we may need more control by specifying the motifs explicitly in a dictionary file, with arbitrary length and IUPAC symbols. This can be facilitated by Topo2Dictfile by first generating a motif dictionary accordingly to user-specified topological patterns, and then editing it as needed, e.g., deleting some motifs, adding more, or introducing IUPAC degeneracy symbols etc.

ProcessFASTA

---

ProcessFASTA is a simple utility program to process a sequence file in FASTA format, e.g., to select a list of sequences based on ids, convert to reverse complementary, combine id and sequence into one-line etc. While such functionalities are surely available in various heavy-duty toolboxes/environments (BioPerl, EMBOSS, BioConductor etc.), none fits ours needs perfectly. We have thus developed this simple utility program mainly for our own convenience.

ProcessTdat

---

This is simple utility program to process a tab-delimited text file, e.g., to extract a subset, perform log transformation, and sort entries by id order etc. It is created following the same idea as for ProcessFASTA.

ExtractWindows

---

A simple utility program to extract sequence fragments from a sequence file, probably of a chromosome.

psamdir2list

---

A Perl utility program to generate a list of PSAM in a given directory. The resultant list can be fed into AffinityProfile or Transfactivity.

AffinityProfile is designed to scan a sequence file against a list of PSAMs or IUPAC motifs to get single base resolution affinity profiles. For convenience, it also allows for a single PSAM (-psam=one_PSAM_file) or an IUPAC motif (-motif=one_IUPAC_motif) to be specified directly on the command line. The PSAMs can be from a MatrixREDUCE or MotifREDUCE run, or a collection of pseudo-PSAMs from literature (as in $REDUCE_SUITE/data/PSAMs/ directory).

AffinityProfile [options] -sequence=seqfile \
                         -psam=one_PSAM_file | -psam_list=list_of_PSAMs |
                         -motif=one_IUPAC_motif | -motif_list=list_of_motifs

  Required parameters:
    -sequence=file_name  --- name of sequence file in FASTA format
    -psam=one_PSAM_file    --- file name of one PSAM
    -psam_list=list_of_PSAMs --- file name containing a list of PSAMs
    -motif=IUPAC_motif     --- one IUPAC motif
    -motif_list=list_of_motifs --- file name containing a list of IUPAC motifs

  Optional parameters:
    [-threshold=float]   --- threshold of affinity for output (0.0)
    [-output=dir_name]   --- path to the output directory (./)
    [-prefix=string]     --- prepended to output profile name (aff_)
    [-affsum=string]     --- file of total affinity per sequence (seq_psam.dat)
    [-detail]            --- also output detailed affinity along each sequence
    [-ids=string]        --- a ',' or ';' delimited list of IDs
    [-column]            --- used with -ids, set profile column-wise for each id
    [-normalize]         --- linear re-scale (per PSAM) the maximum profile to 1.0
    [-strand=integer]      ---  1 |+1 |F | L for leading strand;
                                2 |+2 |B     for both strands;
                               -1 | R |C     for reverse complementary;
                                   with -motif, default to leading strand
                                   with -psam, default to PSAM setting

  Usage:
      (1) AffinityProfile -sequence=$REDUCE_SUITE/data/sequence/YeastUpstream.fasta \
                          -psam_list=psams.list
      (2) AffinityProfile -sequence=$REDUCE_SUITE/data/sequence/YeastUpstream.fasta \
                          -motif=aaaccct
      (3) AffinityProfile -sequence=$REDUCE_SUITE/data/sequence/YeastUpstream.fasta \
                          -motif=aaaga -ids='YAL001C;YAL002W' -column

Notes:

---

• By default, the reported affinities take into account of the threshold (default to 0) specified on the command-line. Thus, per slide window, if the affinity is less than the threshold, it will neither be outputted per window nor added into the sum per sequence (as in file seq_psam_thr.dat, see below).
• AffinityProfile also outputs two files with fixed names: seq_psam.dat (which sums up all affinities, even those below threshold, for reference and comparison) and seq_psam_thr.dat (taking into account of threshold) that contain the sum of affinities per sequence in a tab-delimited format that can be fed directly into Transfactivity.
• Following each run of MatrixREDUCE or MotifREDUCE, two fix-named files, psams.list and motifs.list, are also available, which can be fed into AffinityProfile, as shown in the first example above.
• This is yet another example illustrating how the REDUCE Suite has been designed with tools that are seamlessly inter-connected to allow for great flexibility.

Following a MatrixREDUCE or MotifREDUCE run, a bunch of output files (model parameters, PSAMs etc.) are generated. The HTMLSummary utility program is provided to summarize the main results in a intuitive, easy to follow HTML page so that (even non-expert) users can quickly make sense of their findings, e.g., the top list of significant PSAMs (motifs). Internally, HTMLSummary makes system-calls to the utility program LogoGenerator to create the logo images. These inter-connected programs, plus a few others, constitute the REDUCE Suite.

HTMLSummary [options] [-file=HTMLFile]

  Required parameters:
    none

  Optional parameters:
    [-output=dir_name]  --- path to an MatrixREDUCE/MotifREDUCE run directory,
                            used both as input and output for HTML summary (./)
    [-copy]             --- copy CSS, JavsScript and associated image files to
                            the output directory to make the HTML self-contained
    [-rc]               --- logo based on reverse complementary strand
    [-width=ThumbnailImageWidthInPixel]   (145)
    [-height=ThumbnailImageHeightInPixel] (90)
    [-psam_list=list_of_PSAMs] --- generate a summary LOGO page for the list
                                   of PSAMs in the file
    [-file=HTMLFile]    --- HTML output file name (index.html)

  Usage: (following a MatrixREDUCE run)
    HTMLSummary
    HTMLSummary -psam=psams.list -file=psams.html

REDUCE is an acronym that stands for Regulatory Element Detection Using Correlation with Expression. Based on a simple model for transcriptional regulation by independently acting transcription factors, REDUCE makes it possible to find regulatory elements based on a single microarray experiment. MotifREDUCE in the REDUCE Suite is a more robust and efficient reimplementation of the "original REDUCE algorithm" by Bussemaker et al (2001).

MotifREDUCE [options] -sequence=seqfile -measurement=measfile

  Required parameters:
    -sequence=seqfile     --- sequence file in FASTA format
    -meas=measfile        --- measurement (expression/binding) file in tab-delimited format

  Optional parameters:
    [-topo_list=topofile]  --- name of topology file (up_to_octamers)
    [-topo=topology]       --- single topology pattern, e.g., X3--X4
    [-dicfile=file]        --- list of motifs to check against. IUPAC wild cards
                                   allowed; no length limit
    [-ntop=integer]        --- number of top seed motifs to print out (10)
    [-iupac_pos=integer]   --- number of positions to check for IUPAC degeneracy (0)
    [-iupac_sym=string]    --- IUPAC symbols to check against ('KMRSWYBDHVN')

    [-output=dir_name]     --- path to the output directory (./)
    [-p_value=float]       --- threshold to stop looking for new motifs (0.001)
    [-max_motif=integer]   --- maximum # of motifs to search (20)
    [-strand=integer]      ---  1 |+1 |F | L for leading strand;
                                2 |+2 |B     for both strands;
                               -1 | R |C     for reverse complementary;
                                0 | A |D     auto-detection (check 1 and 2)

    [-runlog=[stderr|stdout|file]]
                           --- direct running diagnostics message to stderr,
                                   stdout or a specific file (stderr)
    [-help]                --- print out this help message

  Usage:
    mkdir -p results   # use topology file (up_to_heptamers)
    MotifREDUCE \
        -meas=$REDUCE_SUITE/examples/MotifREDUCE/yeast_sample.csv \
        -sequence=$REDUCE_SUITE/examples/MotifREDUCE/genome5pns600.fasta \
        -topo_list=$REDUCE_SUITE/examples/MotifREDUCE/up_to_heptamers \
        -o=results
    HTMLSummary -c -o=results

Notes:

---

• The command-line user-interface of MotifREDUCE is identical to that of MatrixREDUCE, but skips the Levenberg-Marquardt non-linear least-squares optimization of weight matrix (Ws). The result is a list of motifs, which are expressed in matrix form with 1s and 0s.
• The above example dataset takes ~10s and finds 10 significant motifs on a contemporary laptop computer.

The REDUCE Suite v2.2 contains a total of 12 programs, as outlined below. The software is distributed with full source code in ANSI C. For the benefit of users, precompiled binaries for the most common Linux, Mac OS X, and Windows operating systems are also provided. Command-line help message is available for each program by specifying the -h option (e.g., LogoGenerator -h), which also include sample usages to get you started. If you have any questions in using the Suite, please do not hesitate to post them on the Forum.

Motif discovery and model building:

---

• MotifREDUCE — An algorithm that builds a motif-based multivariate linear model. REDUCE is an acronym that stands for Regulatory Element Detection Using Correlation with Expression. Based on a simple model for transcriptional regulation by independently acting transcription factors (Bussemaker et al, 2001), REDUCE makes it possible to discover regulatory motifs based on a single microarray experiment. MotifREDUCE is a robust and efficient reimplementation of the original REDUCE algorithm. Required inputs are (i) a genome-wide set of measurements (mRNA expression log-ratios or ChIP fold-enrichments) and (ii) a nucleotide sequence associated with each measurement (e.g., upstream promoter sequence). Output are (i) a set of cis-regulatory oligonucleotide motifs, and (ii) the corresponding regression coefficients.
• MatrixREDUCE — A more sophisticated algorithm that builds a multivariate linear model based on weight matrices (Foat et al., 2005, 2006). Required inputs are the same as far MotifREDUCE: (i) a genome-wide set of measurements (mRNA expression log-ratios or ChIP fold-enrichments) and (ii) a nucleotide sequence associated with each measurement (e.g., upstream promoter sequence). Outputs include (i) the binding specificity, in the form of a position-specific affinity matrix (PSAM), and (ii) the condition-specific concentration/activity for each of a set of trans-acting factors (TF).
• OptimizePSAM — Fits PSAM parameters and coefficients for a single-TF model. MatrixREDUCE makes iterative calls to this program to build a multivariate model.
• Transfactivity — Fit a multivariate linear model to one or more genome-wide sets of measurements. In contrast to MotifREDUCE/MatrixREDUCE, motifs/PSAMs are not inferred from the data, as in, but instead are provided as inputs. This is useful for inferring changes in the (hidden) regulatory activity of one or more TFs of known binding specificity. Transfactivity is a contraction of "trans-factor" and "activity".

Visualization of results:

---

• HTMLSummary — A utility for visualizing the result of a MatrixREDUCE or MotifREDUCE run in HTML format.
• LogoGenerator — A versatile and robust command-line tool that generates logo images in a variety of styles (raw data, frequency, conventional bit information, or affinity logo in ??G). The input can be a PSAM or a multiple sequence alignment file in either FASTA or flat format. The output logo image is in EPS format and is converted to PNG format by default for display in a web page (as from HTMLSummary), using the widely and freely available tool GhostScript tool gs. Other supported image formats include PDF, JPEG, and GIF (further utilizing the convert utility program from ImageMagick).

Affinity-based sequence analysis:

---

• AffinityProfile — Convert one or more DNA/RNA sequences to single-nucleotide resolution affinity profiles or a total regional affinity. A set of motifs and/or PSAMs is required as input.

Miscellaneous utilities:

---

• Convert2PSAM — Convert commonly used motif (pattern) representations of nucleic acid sequences to PSAM format, which is unique to the REDUCE Suite. It also serves to standardize the various formats to a simplified PWM format for easy communication.
• Topo2Dictfile — Generate a motif dictionary file according to user-specified topological patterns, allowing for easy user manipulation (deleting/adding specific motifs, introducing IUPAC degeneracy symbols, etc).
• ProcessFASTA — Process a sequence file in FASTA format to select a list of sequences based on their IDs, convert to reverse complement, combine ID and sequence into a single line, etc.
• ProcessTdat — Manipulate tab-delimited measurement files (extract a subset of experiments, perform log-transformation, sort entries by ID, etc).
• ExtractWindows — Extract subsequences from larger sequences (e.g., a chromosome), based on a set of start/end coordinates.

We are pleased to announce the release of the REDUCE Suite v2.2, a set of software tools to model the regulation of gene expression by transcription factors (TF). By directly correlating genome-wide mRNA expression or TF binding data (e.g. ChIP-chip) with associated nucleotide sequences, the REDUCE Suite can discover the sequence-specific binding affinity of a TF from a single experiment, using all measurements simultaneously, and without using any "background" sequence model.

The REDUCE Suite of software programs has been developed and actively maintained by the Laboratory of Dr. Harmen Bussemaker at the Department of Biological Sciences, Columbia University in the City of New York. The suite has its origin in the REDUCE algorithm of Bussemaker et al. in Nature Genetics (2001), which pioneered the use of motif-based linear regression model to discover cis-regulatory elements (motifs) and infer condition-specific transcription factor activities from a single genome-wide mRNA expression profile. Dr. Barrett Foat, a former graduate student in the Bussemaker Lab, extended REDUCE by adding an optimization procedure to obtain a so-called Position Specific Affinity Matrix (PSAM). He implemented his algorithm in a new program, MatrixREDUCE, using Perl and GNU Scientific Library (GSL). Dr. Xiang-Jun Lu has completely rewritten and significantly enhanced the code using pure ANSI C to make each component program efficient and the whole package self-contained.

Following the release of MatrixREDUCE v1.0 in late 2006, we have made many significant additions and improvements to the software based on extensive feedback from within and outside the lab. Specially, we have greatly improved the calculation of P-values based on a heuristic null model described in Foat et al (2008), and developed a versatile topology-based approach to specify motif patterns. Moreover, we have implemented MotifREDUCE as a standalone, yet more robust and efficient, replacement of the original REDUCE program, and created a command-line driven, general-purpose DNA/RNA-related LogoGenerator. Overall, the suite now consists of more than ten standalone, yet interconnected programs. To better reflect both its root and new versatile functionality, the package has been renamed the "REDUCE Suite", currently at version 2.2.

We understand that getting a scientific software tool published is just the beginning; in the long run, it is the continuous refinements and adaptation to the changing world that make a software suite alive. As a matter of fact, the REDUCE Suite contain many unpublished features. Moreover, while standard "no warranty" applies, we stand firmly behind the software. We strive to get back to your questions, suggestions and bug reports quickly and concretely on the Forum. Browsing the Forum should convince you of our dedication to the REDUCE Suite!

The C source code is in the src/ directory with each tarball. Please refer to the post titled "Set up the REDUCE Suite" on how to compile the code yourself. All REDUCE Suite related questions are welcome on the Forum (only). Do not be shy in sharing openly, but CONCRETELY, any difficult/negative experiences you may have in installing or using the software. By asking your questions on the public Forum, you're benefiting not only yourself but also the user community.

Welcome to the REDUCE Suite, and we look forward to communicating with you on the Forum.

Xiang-Jun Lu & Harmen Bussemaker