Faculty Bibliography

MOTIVATION: Recent advances in sequencing technology have resulted in the dramatic increase of sequencing data, which, in turn, requires efficient management of computational resources, such as computing time, memory requirements as well as prototyping of computational pipelines. RESULTS: We present GenomicTools, a flexible computational platform, comprising both a command-line set of tools and a C++ API, for the analysis and manipulation of high-throughput sequencing data such as DNA-seq, RNA-seq, ChIP-seq and MethylC-seq. GenomicTools implements a variety of mathematical operations between sets of genomic regions thereby enabling the prototyping of computational pipelines that can address a wide spectrum of tasks ranging from pre-processing and quality control to meta-analyses. Additionally, the GenomicTools platform is designed to analyze large datasets of any size by minimizing memory requirements. In practical applications, where comparable, GenomicTools outperforms existing tools in terms of both time and memory usage. AVAILABILITY: The GenomicTools platform (version 2.0.0) was implemented in C++. The source code, documentation, user manual, example datasets and scripts are available online at http://code.google.com/p/ibm-cbc-genomic-tools.

Caveolin-1 (-/-) null stromal cells are a novel genetic model for cancer-associated fibroblasts and myofibroblasts. Here, we used an unbiased informatics analysis of transcriptional gene profiling to show that Cav-1 (-/-) bone-marrow derived stromal cells bear a striking resemblance to the activated tumor stroma of human breast cancers. More specifically, the transcriptional profiles of Cav-1 (-/-) stromal cells were most closely related to the primary tumor stroma of breast cancer patients that had undergone lymph-node (LN) metastasis. This is consistent with previous morphological data demonstrating that a loss of stromal Cav-1 protein (by immuno-histochemical staining in the fibroblast compartment) is significantly associated with increased LN-metastasis. We also provide evidence that the tumor stroma of human breast cancers shows a transcriptional shift towards oxidative stress, DNA damage/repair, inflammation, hypoxia, and aerobic glycolysis, consistent with the "Reverse Warburg Effect". Finally, the tumor stroma of "metastasis-prone" breast cancer patients was most closely related to the transcriptional profiles derived from the brains of patients with Alzheimer's disease. This suggests that certain fundamental biological processes are common to both an activated tumor stroma and neuro-degenerative stress. These processes may include oxidative stress, NO over-production (peroxynitrite formation), inflammation, hypoxia, and mitochondrial dysfunction, which are thought to occur in Alzheimer?s disease pathology. Thus, a loss of Cav-1 expression in cancer-associated myofibroblasts may be a protein biomarker for oxidative stress, aerobic glycolysis, and inflammation, driving the "Reverse Warburg Effect" in the tumor micro-environment and cancer cell metastasis.

The transforming growth factor beta (TGF-beta) family member Decapentaplegic (Dpp) is a key regulator of patterning and growth in Drosophila development. Previous studies have identified a short DNA motif called the silencer element (SE), which recruits a trimeric Smad complex and the repressor Schnurri to downregulate target enhancers upon Dpp signaling. We have now isolated the minimal enhancer of the dad gene and discovered a short motif we termed the activating element (AE). The AE is similar to the SE and recruits the Smad proteins via a conserved mechanism. However, the AE and SE differ at important nucleotide positions. As a consequence, the AE does not recruit Schnurri but rather integrates repressive input by the default repressor Brinker and activating input by the Smad signal transducers Mothers against Dpp (Mad) and Medea via competitive DNA binding. The AE allows the identification of hitherto unknown direct Dpp targets and is functionally conserved in vertebrates.

Alu and B1 repeats are mobile elements that originated in an initial duplication of the 7SL RNA gene prior to the primate-rodent split about 80 million years ago and currently account for a substantial fraction of the human and mouse genome, respectively. Following the primate-rodent split, Alu and B1 elements spread independently in each of the two genomes in a seemingly random manner, and, according to the prevailing hypothesis, negative selection shaped their final distribution in each genome by forcing the selective loss of certain Alu and B1 copies. In this paper, contrary to the prevailing hypothesis, we present evidence that Alu and B1 elements have been selectively retained in the upstream and intronic regions of genes belonging to specific functional classes. At the same time, we found no evidence for selective loss of these elements in any functional class. A subset of the functional links we discovered corresponds to functions where Alu involvement has actually been experimentally validated, whereas the majority of the functional links we report are novel. Finally, the unexpected finding that Alu and B1 elements show similar biases in their distribution across functional classes, despite having spread independently in their respective genomes, further supports our claim that the extant instances of Alu and B1 elements are the result of positive selection.

The Bicoid (Bcd) transcription factor is distributed as a long-range concentration gradient along the anterior posterior (AP) axis of the Drosophila embryo. Bcd is required for the activation of a series of target genes, which are expressed at specific positions within the gradient. Here we directly tested whether different concentration thresholds within the Bcd gradient establish the relative positions of its target genes by flattening the gradient and systematically varying expression levels. Genome-wide expression profiles were used to estimate the total number of Bcd target genes, and a general correlation was found between the Bcd concentration required for activation and the positions where target genes are expressed in wild-type embryos. However, concentrations required for target gene activation in embryos with flattened Bcd were consistently lower than those present at each target gene's position in the wild-type gradient, suggesting that Bcd is in excess at every position along the AP axis. Also, several Bcd target genes were positioned in correctly ordered stripes in embryos with flattened Bcd, and we suggest that these stripes are normally regulated by interactions between Bcd and the terminal patterning system. Our findings argue strongly against the strict interpretation of the Bcd morphogen hypothesis, and support the idea that target gene positioning involves combinatorial interactions that are mediated by the binding site architecture of each gene's cis-regulatory elements.

We identified the most frequent, variable-length DNA sequence motifs in the human and mouse genomes and sub-selected those with multiple recurrences in the intergenic and intronic regions and at least one additional exonic instance in the corresponding genome. We discovered that these motifs have virtually no overlap with intronic sequences that are conserved between human and mouse, and thus are genome-specific. Moreover, we found that these motifs span a substantial fraction of previously uncharacterized human and mouse intronic space. Surprisingly, we found that these genome-specific motifs are over-represented in the introns of genes belonging to the same biological processes and molecular functions in both the human and mouse genomes even though the underlying sequences are not conserved between the two genomes. In fact, the processes and functions that are linked to these genome-specific sequence-motifs are distinct from the processes and functions which are associated with intronic regions that are conserved between human and mouse. The findings show that intronic regions from different genomes are linked to the same processes and functions in the absence of underlying sequence conservation. We highlight the ramifications of this observation with a concrete example that involves the microsatellite instability gene MLH1.

In earlier work, we introduced and discussed a generalized computational framework for identifying horizontal transfers. This framework relied on a gene's nucleotide composition, obviated the need for knowledge of codon boundaries and database searches, and was shown to perform very well across a wide range of archaeal and bacterial genomes when compared with previously published approaches, such as Codon Adaptation Index and C + G content. Nonetheless, two considerations remained outstanding: we wanted to further increase the sensitivity of detecting horizontal transfers and also to be able to apply the method to increasingly smaller genomes. In the discussion that follows, we present such a method, Wn-SVM, and show that it exhibits a very significant improvement in sensitivity compared with earlier approaches. Wn-SVM uses a one-class support-vector machine and can learn using rather small training sets. This property makes Wn-SVM particularly suitable for studying small-size genomes, similar to those of viruses, as well as the typically larger archaeal and bacterial genomes. We show experimentally that the new method results in a superior performance across a wide range of organisms and that it improves even upon our own earlier method by an average of 10% across all examined genomes. As a small-genome case study, we analyze the genome of the human cytomegalovirus and demonstrate that Wn-SVM correctly identifies regions that are known to be conserved and prototypical of all beta-herpesvirinae, regions that are known to have been acquired horizontally from the human host and, finally, regions that had not up to now been suspected to be horizontally transferred. Atypical region predictions for many eukaryotic viruses, including the alpha-, beta- and gamma-herpesvirinae, and 123 archaeal and bacterial genomes, have been made available online at http://cbcsrv.watson.ibm.com/HGT_SVM/.

In recent years, the increase in the amounts of available genomic data has made it easier to appreciate the extent by which organisms increase their genetic diversity through horizontally transferred genetic material. Such transfers have the potential to give rise to extremely dynamic genomes where a significant proportion of their coding DNA has been contributed by external sources. Because of the impact of these horizontal transfers on the ecological and pathogenic character of the recipient organisms, methods are continuously sought that are able to computationally determine which of the genes of a given genome are products of transfer events. In this paper, we introduce and discuss a novel computational method for identifying horizontal transfers that relies on a gene's nucleotide composition and obviates the need for knowledge of codon boundaries. In addition to being applicable to individual genes, the method can be easily extended to the case of clusters of horizontally transferred genes. With the help of an extensive and carefully designed set of experiments on 123 archaeal and bacterial genomes, we demonstrate that the new method exhibits significant improvement in sensitivity when compared to previously published approaches. In fact, it achieves an average relative improvement across genomes of between 11 and 41% compared to the Codon Adaptation Index method in distinguishing native from foreign genes. Our method's horizontal gene transfer predictions for 123 microbial genomes are available online at http://cbcsrv.watson.ibm.com/HGT/.