Search tips
Search criteria

Results 1-19 (19)

Clipboard (0)

Select a Filter Below

more »
Year of Publication
Document Types
1.  Targeted and genome-wide sequencing reveal single nucleotide variations impacting specificity of Cas9 in human stem cells 
Nature Communications  2014;5:5507.
CRISPR/Cas9 has demonstrated a high-efficiency in site-specific gene targeting. However, potential off-target effects of the Cas9 nuclease represent a major safety concern for any therapeutic application. Here, we knock out the Tafazzin gene by CRISPR/Cas9 in human-induced pluripotent stem cells with 54% efficiency. We combine whole-genome sequencing and deep-targeted sequencing to characterise the off-target effects of Cas9 editing. Whole-genome sequencing of Cas9-modified hiPSC clones detects neither gross genomic alterations nor elevated mutation rates. Deep sequencing of in silico predicted off-target sites in a population of Cas9-treated cells further confirms high specificity of Cas9. However, we identify a single high-efficiency off-target site that is generated by a common germline single-nucleotide variant (SNV) in our experiment. Based on in silico analysis, we estimate a likelihood of SNVs creating off-target sites in a human genome to be ~1.5–8.5%, depending on the genome and site-selection method, but also note that mutations might be generated at these sites only at low rates and may not have functional consequences. Our study demonstrates the feasibility of highly specific clonal ex vivo gene editing using CRISPR/Cas9 and highlights the value of whole-genome sequencing before personalised CRISPR design.
The microbial RNA-guided CRISPR/Cas9 system has robust genome-editing activities, but the off-target effects of the Cas9 nuclease have only recently begun to be analysed. Here the authors provide evidence for high specificity of the Cas9 nuclease on targeting of the Tafazzin gene in human-induced pluripotent stem cells and demonstrate the impact of single-nucleotide variations of the human genome on Cas9 specificity.
PMCID: PMC4352754  PMID: 25425480
2.  Multiplex single-molecule interaction profiling of DNA barcoded proteins 
Nature  2014;515(7528):554-557.
In contrast with advances in massively parallel DNA sequencing1, high-throughput protein analyses2-4 are often limited by ensemble measurements, individual analyte purification and hence compromised quality and cost-effectiveness. Single-molecule (SM) protein detection achieved using optical methods5 is limited by the number of spectrally nonoverlapping chromophores. Here, we introduce a single molecular interaction-sequencing (SMI-Seq) technology for parallel protein interaction profiling leveraging SM advantages. DNA barcodes are attached to proteins collectively via ribosome display6 or individually via enzymatic conjugation. Barcoded proteins are assayed en masse in aqueous solution and subsequently immobilized in a polyacrylamide (PAA) thin film to construct a random SM array, where barcoding DNAs are amplified into in situ polymerase colonies (polonies)7 and analyzed by DNA sequencing. This method allows precise quantification of various proteins with a theoretical maximum array density of over one million polonies per square millimeter. Furthermore, protein interactions can be measured based on the statistics of colocalized polonies arising from barcoding DNAs of interacting proteins. Two demanding applications, G-protein coupled receptor (GPCR) and antibody binding profiling, were demonstrated. SMI-Seq enables “library vs. library” screening in a one-pot assay, simultaneously interrogating molecular binding affinity and specificity.
PMCID: PMC4246050  PMID: 25252978
3.  Multi-kilobase homozygous targeted gene replacement in human induced pluripotent stem cells 
Nucleic Acids Research  2014;43(3):e21.
Sequence-specific nucleases such as TALEN and the CRISPR/Cas9 system have so far been used to disrupt, correct or insert transgenes at precise locations in mammalian genomes. We demonstrate efficient ‘knock-in’ targeted replacement of multi-kilobase genes in human induced pluripotent stem cells (iPSC). Using a model system replacing endogenous human genes with their mouse counterpart, we performed a comprehensive study of targeting vector design parameters for homologous recombination. A 2.7 kilobase (kb) homozygous gene replacement was achieved in up to 11% of iPSC without selection. The optimal homology arm length was around 2 kb, with homology length being especially critical on the arm not adjacent to the cut site. Homologous sequence inside the cut sites was detrimental to targeting efficiency, consistent with a synthesis-dependent strand annealing (SDSA) mechanism. Using two nuclease sites, we observed a high degree of gene excisions and inversions, which sometimes occurred more frequently than indel mutations. While homozygous deletions of 86 kb were achieved with up to 8% frequency, deletion frequencies were not solely a function of nuclease activity and deletion size. Our results analyzing the optimal parameters for targeting vector design will inform future gene targeting efforts involving multi-kilobase gene segments, particularly in human iPSC.
PMCID: PMC4330342  PMID: 25414332
4.  Highly multiplexed subcellular RNA sequencing in situ 
Science (New York, N.Y.)  2014;343(6177):1360-1363.
Understanding the spatial organization of gene expression with single nucleotide resolution requires localizing the sequences of expressed RNA transcripts within a cell in situ. Here we describe fluorescent in situ RNA sequencing (FISSEQ), in which stably cross-linked cDNA amplicons are sequenced within a biological sample. Using 30-base reads from 8,742 genes in situ, we examined RNA expression and localization in human primary fibroblasts using a simulated wound healing assay. FISSEQ is compatible with tissue sections and whole mount embryos, and reduces the limitations of optical resolution and noisy signals on single molecule detection. Our platform enables massively parallel detection of genetic elements, including gene transcripts and molecular barcodes, and can be used to investigate cellular phenotype, gene regulation, and environment in situ.
PMCID: PMC4140943  PMID: 24578530
5.  Improved Cell-Free RNA and Protein Synthesis System 
PLoS ONE  2014;9(9):e106232.
Cell-free RNA and protein synthesis (CFPS) is becoming increasingly used for protein production as yields increase and costs decrease. Advances in reconstituted CFPS systems such as the Protein synthesis Using Recombinant Elements (PURE) system offer new opportunities to tailor the reactions for specialized applications including in vitro protein evolution, protein microarrays, isotopic labeling, and incorporating unnatural amino acids. In this study, using firefly luciferase synthesis as a reporter system, we improved PURE system productivity up to 5 fold by adding or adjusting a variety of factors that affect transcription and translation, including Elongation factors (EF-Ts, EF-Tu, EF-G, and EF4), ribosome recycling factor (RRF), release factors (RF1, RF2, RF3), chaperones (GroEL/ES), BSA and tRNAs. The work provides a more efficient defined in vitro transcription and translation system and a deeper understanding of the factors that limit the whole system efficiency.
PMCID: PMC4152126  PMID: 25180701
7.  CAS9 transcriptional activators for target specificity screening and paired nickases for cooperative genome engineering 
Nature biotechnology  2013;31(9):10.1038/nbt.2675.
Prokaryotic type II CRISPR-Cas systems can be adapted to enable targeted genome modifications across a range of eukaryotes.1–7. Here we engineer this system to enable RNA-guided genome regulation in human cells by tethering transcriptional activation domains either directly to a nuclease-null Cas9 protein or to an aptamer-modified single guide RNA (sgRNA). Using this functionality we developed a novel transcriptional activation–based assay to determine the landscape of off-target binding of sgRNA:Cas9 complexes and compared it with the off-target activity of transcription activator–like (TAL) effector proteins8, 9. Our results reveal that specificity profiles are sgRNA dependent, and that sgRNA:Cas9 complexes and 18-mer TAL effector proteins can potentially tolerate 1–3 and 1–2 target mismatches, respectively. By engineering a requirement for cooperativity through offset nicking for genome editing or through multiple synergistic sgRNAs for robust transcriptional activation, we suggest methods to mitigate off-target phenomena. Our results expand the versatility of the sgRNA:Cas9 tool and highlight the critical need to engineer improved specificity.
PMCID: PMC3818127  PMID: 23907171
8.  Rational optimization of tolC as a powerful dual selectable marker for genome engineering 
Nucleic Acids Research  2014;42(7):4779-4790.
Selection has been invaluable for genetic manipulation, although counter-selection has historically exhibited limited robustness and convenience. TolC, an outer membrane pore involved in transmembrane transport in E. coli, has been implemented as a selectable/counter-selectable marker, but counter-selection escape frequency using colicin E1 precludes using tolC for inefficient genetic manipulations and/or with large libraries. Here, we leveraged unbiased deep sequencing of 96 independent lineages exhibiting counter-selection escape to identify loss-of-function mutations, which offered mechanistic insight and guided strain engineering to reduce counter-selection escape frequency by ∼40-fold. We fundamentally improved the tolC counter-selection by supplementing a second agent, vancomycin, which reduces counter-selection escape by 425-fold, compared colicin E1 alone. Combining these improvements in a mismatch repair proficient strain reduced counter-selection escape frequency by 1.3E6-fold in total, making tolC counter-selection as effective as most selectable markers, and adding a valuable tool to the genome editing toolbox. These improvements permitted us to perform stable and continuous rounds of selection/counter-selection using tolC, enabling replacement of 10 alleles without requiring genotypic screening for the first time. Finally, we combined these advances to create an optimized E. coli strain for genome engineering that is ∼10-fold more efficient at achieving allelic diversity than previous best practices.
PMCID: PMC3985617  PMID: 24452804
9.  Barcoding cells using cell-surface programmable DNA-binding domains 
Nature methods  2013;10(5):403-406.
We develop here a novel approach to barcode large numbers of cells through cell-surface expression of programmable zinc-finger DNA-binding domains (sZFs). We show sZFs enable double-stranded DNA to sequence-specifically label living cells, and also develop a sequential tagging approach to in situ image >3 cell types using just 3 fluorophores. Finally we demonstrate their broad versatility through ability to serve as surrogate reporters and facilitate selective cell capture and targeting.
PMCID: PMC3641172  PMID: 23503053
10.  On the design of clone-based haplotyping 
Genome Biology  2013;14(9):R100.
Haplotypes are important for assessing genealogy and disease susceptibility of individual genomes, but are difficult to obtain with routine sequencing approaches. Experimental haplotype reconstruction based on assembling fragments of individual chromosomes is promising, but with variable yields due to incompletely understood parameter choices.
We parameterize the clone-based haplotyping problem in order to provide theoretical and empirical assessments of the impact of different parameters on haplotype assembly. We confirm the intuition that long clones help link together heterozygous variants and thus improve haplotype length. Furthermore, given the length of the clones, we address how to choose the other parameters, including number of pools, clone coverage and sequencing coverage, so as to maximize haplotype length. We model the problem theoretically and show empirically the benefits of using larger clones with moderate number of pools and sequencing coverage. In particular, using 140 kb BAC clones, we construct haplotypes for a personal genome and assemble haplotypes with N50 values greater than 2.6 Mb. These assembled haplotypes are longer and at least as accurate as haplotypes of existing clone-based strategies, whether in vivo or in vitro.
Our results provide practical guidelines for the development and design of clone-based methods to achieve long range, high-resolution and accurate haplotypes.
PMCID: PMC4053695  PMID: 24028704
11.  RNA-Guided Human Genome Engineering via Cas9 
Science (New York, N.Y.)  2013;339(6121):823-826.
PMCID: PMC3712628  PMID: 23287722
12.  Optimization of scarless human stem cell genome editing 
Nucleic Acids Research  2013;41(19):9049-9061.
Efficient strategies for precise genome editing in human-induced pluripotent cells (hiPSCs) will enable sophisticated genome engineering for research and clinical purposes. The development of programmable sequence-specific nucleases such as Transcription Activator-Like Effectors Nucleases (TALENs) and Cas9-gRNA allows genetic modifications to be made more efficiently at targeted sites of interest. However, many opportunities remain to optimize these tools and to enlarge their spheres of application. We present several improvements: First, we developed functional re-coded TALEs (reTALEs), which not only enable simple one-pot TALE synthesis but also allow TALE-based applications to be performed using lentiviral vectors. We then compared genome-editing efficiencies in hiPSCs mediated by 15 pairs of reTALENs and Cas9-gRNA targeting CCR5 and optimized ssODN design in conjunction with both methods for introducing specific mutations. We found Cas9-gRNA achieved 7–8× higher non-homologous end joining efficiencies (3%) than reTALENs (0.4%) and moderately superior homology-directed repair efficiencies (1.0 versus 0.6%) when combined with ssODN donors in hiPSCs. Using the optimal design, we demonstrated a streamlined process to generated seamlessly genome corrected hiPSCs within 3 weeks.
PMCID: PMC3799423  PMID: 23907390
13.  Genome engineering in Saccharomyces cerevisiae using CRISPR-Cas systems 
Nucleic Acids Research  2013;41(7):4336-4343.
Clustered Regularly Interspaced Short Palindromic Repeats (CRISPR) and CRISPR-associated (Cas) systems in bacteria and archaea use RNA-guided nuclease activity to provide adaptive immunity against invading foreign nucleic acids. Here, we report the use of type II bacterial CRISPR-Cas system in Saccharomyces cerevisiae for genome engineering. The CRISPR-Cas components, Cas9 gene and a designer genome targeting CRISPR guide RNA (gRNA), show robust and specific RNA-guided endonuclease activity at targeted endogenous genomic loci in yeast. Using constitutive Cas9 expression and a transient gRNA cassette, we show that targeted double-strand breaks can increase homologous recombination rates of single- and double-stranded oligonucleotide donors by 5-fold and 130-fold, respectively. In addition, co-transformation of a gRNA plasmid and a donor DNA in cells constitutively expressing Cas9 resulted in near 100% donor DNA recombination frequency. Our approach provides foundations for a simple and powerful genome engineering tool for site-specific mutagenesis and allelic replacement in yeast.
PMCID: PMC3627607  PMID: 23460208
14.  Proteome-wide systems analysis of a cellulosic biofuel-producing microbe 
We apply mass spectrometry-based ReDi proteomics to quantify the Clostridium phytofermentans proteome during fermentation of cellulosic substrates. ReDi proteomics gives accurate, low-cost quantification of an extra and intracellular microbial proteome. When combined with physiological measurements, these methods form a general systems biology strategy to evaluate the efficiency of cellulosic bioconversion and to identify enzyme targets to engineer for improving this process.C. phytofermentans expressed more than 100 carbohydrate-active enzymes, of which distinct subsets were upregulated on cellulose and hemicellulose. Numerous extracellular enzymes cleave insoluble plant polysaccharides into oligosaccharides, which are transported into the cell to be further degraded by intracellular carbohydratases. Sugars are catabolized by EMP glycolysis incorporating alternative glycolytic enzymes to maximize the ATP yield of anaerobic metabolism.During cellulosic fermentation, cells adhered to the substrate and altered metabolic processes such as upregulation of tryptophan and nicotinamide synthesis proteins and repression of proteins for fatty acid metabolism and cell motility. These diverse metabolic changes highlight how a systems approach can identify novel ways to optimize cellulosic fermentation.
Cellulose is the world's most abundant renewable, biological energy source (Leschine, 1995). Microbial fermentation of cellulosic biomass could sustainably provide enough ethanol for 65% of US ground transportation fuel at current levels (Somerville, 2006). However, cellulose in plant biomass is packaged into a crystalline matrix, making biomass deconstruction a key roadblock to using it as a feedstock (Houghton et al, 2006). A promising strategy to overcome biomass recalcitrance is consolidated bioprocessing (Lynd et al, 2002), which uses microbes such as Clostridium phytofermentans to both secrete enzymes to depolymerize biomass and then ferment the resulting hexose and pentose sugars to a biofuel such as ethanol. The C. phytofermentans genome encodes 161 carbohydrate-active enzymes (CAZy) including 108 glycoside hydrolases spread across 39 families (Cantarel et al, 2009), highlighting the elaborate set of enzymes needed to breakdown different cellulosic polysaccharides. Faced with the complexity of metabolizing biomass, systems biology strategies are needed to comprehensively identify which cellulolytic and metabolic enzymes are used to ferment different cellulosic substrates.
This study presents a systems-level analysis of how C. phytofermentans ferments different cellulosic substrates that incorporates quantitative mass spectrometry-based proteomics of over 2500 proteins. Protein concentrations within each carbon source treatment were calculated by machine learning-supported spectral counting (Absolute Protein EXpression, APEX) (Lu et al, 2007). Protein levels on hemicellulose and cellulose relative to glucose were determined using reductive methylation (Hsu et al, 2003; Boersema et al, 2009), here called ReDi labeling, to chemically incorporate hydrogen or deuterium isotopes at lysines and N-terminal amines of tryptic peptides. We show that ReDi proteomics gives accurate, low-cost quantification of a microbial proteome and can be used to discern extracellular proteins. Further, we combine these quantitative proteomics with detailed measurements of growth, biomass consumption, fermentation product analyses, and electron microscopy. Together, these methods form a general strategy to evaluate the efficiency of cellulosic bioconversion and to identify enzyme targets to engineer for improving this process (Figure 1).
We found that fermentation of cellulosic substrates by C. phytofermentans involves secretion of numerous CAZy as well as proteins for binding of extracellular solutes, proteolysis, and motility. The most highly expressed protein in the proteome is a secreted protein that appears to compose a surface layer to support the cell and anchor cell surface proteins, including some enzymes for plant degradation. Once the secreted CAZy cleave insoluble plant polysaccharides into oligosaccharides, they are taken into the cell to be further degraded by intracellular CAZy, enabling more efficient sugar transport, conserving energy by phosphorolytic cleavage, and ensuring the sugar monomers were not available to competing microbes. Sugars are catabolized by EMP glycolysis incorporating reversible, PPi-dependent glycolytic enzymes, and pyruvate ferredoxin oxidoreductase. The genome encodes seven alcohol dehydrogenases, among which two iron-dependent enzymes are highly expressed and likely facilitate the high ethanol yields. Growth on cellulose also resulted in indirect changes such as increased tryptophan and nicotinamide synthesis and repression of fatty acid synthesis. We distilled the data into a model showing the highly expressed enzymes enabling efficient cellulosic fermentation by C. phytofermentans (Figure 7). Collectively, these data help understand how bacteria recycle plant biomass works towards enabling the use of plant biomass as a low-cost chemical feedstock.
Fermentation of plant biomass by microbes like Clostridium phytofermentans recycles carbon globally and can make biofuels from inedible feedstocks. We analyzed C. phytofermentans fermenting cellulosic substrates by integrating quantitative mass spectrometry of more than 2500 proteins with measurements of growth, enzyme activities, fermentation products, and electron microscopy. Absolute protein concentrations were estimated using Absolute Protein EXpression (APEX); relative changes between treatments were quantified with chemical stable isotope labeling by reductive dimethylation (ReDi). We identified the different combinations of carbohydratases used to degrade cellulose and hemicellulose, many of which were secreted based on quantification of supernatant proteins, as well as the repertoires of glycolytic enzymes and alcohol dehydrogenases (ADHs) enabling ethanol production at near maximal yields. Growth on cellulose also resulted in diverse changes such as increased expression of tryptophan synthesis proteins and repression of proteins for fatty acid metabolism and cell motility. This study gives a systems-level understanding of how this microbe ferments biomass and provides a rational, empirical basis to identify engineering targets for industrial cellulosic fermentation.
PMCID: PMC3049413  PMID: 21245846
bioenergy; clostridium; proteomics
15.  Personal genomes in progress: from the Human Genome Project to the Personal Genome Project 
The cost of a diploid human genome sequence has dropped from about $70M to $2000 since 2007- even as the standards for redundancy have increased from 7x to 40x in order to improve call rates. Coupled with the low return on investment for common single-nucleotide polymorphisms, this has caused a significant rise in interest in correlating genome sequences with comprehensive environmental and trait data (GET). The cost of electronic health records, imaging, and microbial, immunological, and behavioral data are also dropping quickly. Sharing such integrated GET datasets and their interpretations with a diversity of researchers and research subjects highlights the need for informed-consent models capable of addressing novel privacy and other issues, as well as for flexible data-sharing resources that make materials and data available with minimum restrictions on use. This article examines the Personal Genome Project's effort to develop a GET database as a public genomics resource broadly accessible to both researchers and research participants, while pursuing the highest standards in research ethics.
PMCID: PMC3181947  PMID: 20373666
Personal Genome Project; personal genomics; DNA sequencing technology; whole-genome sequencing; phenome; envirome; microbiome; GET data set; open consent; public genome; ELSI
16.  Digital RNA Allelotyping Reveals Tissue-specific and Allele-specific Gene Expression in Human 
Nature methods  2009;6(8):613-618.
We developed a digital RNA allelotyping method for quantitatively interrogating allele-specific gene expression. This method involves ultra-deep sequencing of padlock captured SNPs from the transcriptome. We characterized four cell lines established from two human subjects in the Personal Genome Project. Approximately 11–22% of the heterozygous mRNA-associated SNPs show allele-specific expression in each cell line; and 4.3–8.5% are tissue-specific, suggesting the presence of tissue-specific cis-regulation. When applied to two pairs of sibling human embryonic stem cell lines, the sibling lines were more similar in allele-specific expression than were the genetically unrelated lines. We found that the variation of allelic ratios in gene expression among different cell lines is primarily explained by genetic variations, much more so than by specific tissue types or culturing conditions. Comparison of expressed SNPs on the sense and anti-sense transcripts suggested that allelic ratios are primarily determined by cis-regulatory mechanisms on the sense transcripts.
PMCID: PMC2742772  PMID: 19620972
17.  Identification of molecular markers of bipolar cells in the murine retina 
The Journal of Comparative Neurology  2008;507(5):1795-1810.
Retinal bipolar neurons serve as relay interneurons that connect rod and cone photoreceptor cells to amacrine and ganglion cells. They exhibit diverse morphologies essential for correct routing of photoreceptor cell signals to specific postsynaptic amacrine and ganglion cells. The development and physiology of these interneurons have not been completely defined molecularly. Despite previous identification of genes expressed in several bipolar cell subtypes, molecules that mark each bipolar cell type still await discovery. In this report, novel genetic markers of murine bipolar cells were found. Candidates were initially generated by using microarray analysis of single bipolar cells and mining of retinal serial analysis of gene expression (SAGE) data. These candidates were subsequently tested for expression in bipolar cells by RNA in situ hybridization. Ten new molecular markers were identified, five of which are highly enriched in their expression in bipolar cells within the adult retina. Double-labeling experiments using probes for previously characterized subsets of bipolar cells were performed to identify the subtypes of bipolar cells that express the novel markers. Additionally, the expression of bipolar cell genes was analyzed in Bhlhb4 knockout retinas, in which rod bipolar cells degenerate postnatally, to delineate further the identity of bipolar cells in which novel markers are found. From the analysis of Bhlhb4 mutant retinas, cone bipolar cell gene expression appears to be relatively unaffected by the degeneration of rod bipolar cells. Identification of molecular markers for the various subtypes of bipolar cells will lead to greater insights into the development and function of these diverse interneurons.
PMCID: PMC2665264  PMID: 18260140
retina; bipolar cells; Bhlhb4; gene expression; microarray; mouse
18.  Global gene expression of Prochlorococcus ecotypes in response to changes in nitrogen availability 
Nitrogen (N) often limits biological productivity in the oceanic gyres where Prochlorococcus is the most abundant photosynthetic organism. The Prochlorococcus community is composed of strains, such as MED4 and MIT9313, that have different N utilization capabilities and that belong to ecotypes with different depth distributions. An interstrain comparison of how Prochlorococcus responds to changes in ambient nitrogen is thus central to understanding its ecology. We quantified changes in MED4 and MIT9313 global mRNA expression, chlorophyll fluorescence, and photosystem II photochemical efficiency (Fv/Fm) along a time series of increasing N starvation. In addition, the global expression of both strains growing in ammonium-replete medium was compared to expression during growth on alternative N sources. There were interstrain similarities in N regulation such as the activation of a putative NtcA regulon during N stress. There were also important differences between the strains such as in the expression patterns of carbon metabolism genes, suggesting that the two strains integrate N and C metabolism in fundamentally different ways.
PMCID: PMC1682016  PMID: 17016519
cyanobacteria; interstrain; nitrogen; Prochlorococcus; transcription
19.  Automated modelling of signal transduction networks 
BMC Bioinformatics  2002;3:34.
Intracellular signal transduction is achieved by networks of proteins and small molecules that transmit information from the cell surface to the nucleus, where they ultimately effect transcriptional changes. Understanding the mechanisms cells use to accomplish this important process requires a detailed molecular description of the networks involved.
We have developed a computational approach for generating static models of signal transduction networks which utilizes protein-interaction maps generated from large-scale two-hybrid screens and expression profiles from DNA microarrays. Networks are determined entirely by integrating protein-protein interaction data with microarray expression data, without prior knowledge of any pathway intermediates. In effect, this is equivalent to extracting subnetworks of the protein interaction dataset whose members have the most correlated expression profiles.
We show that our technique accurately reconstructs MAP Kinase signaling networks in Saccharomyces cerevisiae. This approach should enhance our ability to model signaling networks and to discover new components of known networks. More generally, it provides a method for synthesizing molecular data, either individual transcript abundance measurements or pairwise protein interactions, into higher level structures, such as pathways and networks.
PMCID: PMC137599  PMID: 12413400

Results 1-19 (19)