School of Medical Sciences

Bioinformatics of disease

Principal Investigator (PI)


Cris Print's laboratory focuses on using bioinformatic methods and integrative biology methods to understand disease.

A major challenge facing medical science at present is the integration of vast and rapidly growing volumes of information into a holistic understanding of disease. In some cases this integration involves the generation of mathematical models such as gene regulatory networks - in other cases information can be integrated using simple and common-sense methods. Information that may be informative about disease processes includes:

  • mRNA and miRNA information from microarrays or RNAseq
  • genotype information from deep sequencing or SNP analysis
  • clinical information
  • digitised pathological information
  • information from cell biology and transgenic experiments.

Examples of research from our laboratory that attempt to integrate information of different types synergistically to understand disease are detailed below and cover the following general areas:

Some of this work is based in The University of Auckland, while other projects are based in the Auckland Bioengineering Institute, Dr Andrew Shelling's Medical Genetics Research Group, the biotechnology company GNI Ltd, and Cambridge University in the UK, where Cris worked until mid 2005.

For general enquiries about this research please email Cris Print.


Lab members

  • Cherie Blenkiron
  • Daniel Hurley (Auckland Bioengineering Institute)
  • Annette Lasham
  • Sunali Mehta
  • Andrew Miller (Auckland Bioengineering Institute)
  • Anita Muthukaruppan (Department of Obstetrics and Gynaecology)
  • Li Wang
  • Wendy Watkins
  • Edward Walker (at The Plant and Food Institute)
  • Deborah Wright

Current NZ collaborators/co-PIs

Current overseas collaborators

  • Hiromitsu Araki (GNI Ltd)
  • Louise Hull and Maria Ohlsson-Teague (Adelaide University)
  • Satoru Miyano, Seiya Imoto and Yoshi Tamada (Tokyo University)
  • Stephen Charnock-Jones (Cambridge University)

The lab's ethos

For this work to be successful we believe that several points are important.

  • People. These multidisciplinary projects require good collaboration between specialists in compuational biology, applied mathematics, statistics, genetics, and most importantly with pathologists, physicians and surgeons. In addition to these specialists, to coordinate the projects it is essential  to have generalists with a broad understanding of medicine, cellular and molecular biology, statistics and computational biology. Postgraduate students in our laboratory are exposed to several research disciplines and gain experience in collaborating and communicating across traditional research boundaries.
  • Conservative analysis. It seems most productive to minimise the chances of "false discovery" by focussing on the intersection of different types of information, and by setting a high threshold  for accepting our hypotheses. Methods such as Bayesian models in which uncertainty can be explicitly modelled are very attractive.
  • Validity. We do not get much satisfaction from simply combining high content information from different sources, or from simply generating  mathematical models of the molecular aspects of a disease, no matter how elegant the methods used are! Evaluation of this type of research by laboratory experiments or clinical trials is important if we are to translate high content information into improved biological understanding or clinical practice.

General examples of our research

(1) Endothelial cell biology

Endothelial cells line blood vessels and orchestrate the growth and regression of blood vessels to meet the changing demands of the tissues they supply. We have studied how endothelial cells contribute to blood vessel biology and pathology using a number of techniques. We have focused particularly on the process of endothelial cell apoptosis (programmed cell suicide). We have studied the role of apoptosis in blood vessel development using in vitro assays of angiogenesis, in which human endothelial cells "grow" into blood vessels in tissue culture. Apoptosis occurs as the blood vessels grow (Figure 1, Duval et. al. Angiogenesis; 6; 171-183, 2003).  


Figure 1. Apoptosis (measured by caspase 3 activation, orange) occurs in the vessel-like structures as they develop (click image to see higher resolution version).

We have used tissue-specific transgenes to turn off the process of apoptosis specifically in mouse endothelial cells. This has revealed that endothelial cell apoptosis is essential for blood vessel development in embryos, since without it embryos die before birth and show a multitude of developmental defects in small blood vessels while maintaining normal development of large blood vessels (Figure 2, Duval et. al. Angiogenesis; 10; 55-68, 2007).


Figure 2. Placental blood vessels (brown) fail to develop in transgenic mouse embryos in which endothelial cell apoptosis has been inactivated. WT = wild type, TG = transgenic, F = foetal capillary, MS = maternal sinus and H = haemorrhage (click image to see higher resolution version).

We are currently using transgenic zebrafish to further study the role played endothelial cell apoptosis in vascular development, in collaboration with A/Prof Philip Crosier and Dr Maria Vega Flores.

Endothelial cell apoptosis is thought to be important in matching vessel anatomy to tissue requirements for blood supply, and for patterning the cardiovascular system during development, as well as for some types of cancer chemotherapy (stromal targeting drugs).  It is driven by an orchestrated "suicide program", which includes protein-based (proteome), gene expression-based (transcriptome) and cell surface carbohydrate-based (glycome) events. We have used gene arrays and glycomics to map some of the events that may contribute to the process of endothelial cell apoptosis (figure 3, FASEB J 18; 188-190, 2003). Using a similar approach, we collaborated with Professor Jordan Pober from Yale University in the USA to understand how the apoptosis regulator Bcl-2 may influence blood vessel maturation (Endothelium, 15(1); 59-71, 2008).


Figure 3. Summary of the potential roles that regulated RNA transcript abundance and carbohydrate sulphation may play in preparing cells for apoptosis. Abbreviations; ECM, extra-cellular matrix; HSGAG, heparan sulfate glycosaminoglycan (click image to see higher resolution version).

We have used Bayesian Gene Network techniques in an attempt to infer the genetic networks that may underlie the process of apoptosis in endothelial cells (figure 4). One of the master regulators of apoptosis hypothesised by these networks (GABARAP) appears to be critical for apoptosis, since when its expression is knocked down using RNAi, endothelial cell apoptosis is strongly inhibited (Affara et. al. Philosophical; Transactions of the Royal Society; 362; 1469–1487, 2007).


Figure 4. A graph representing a dynamic Bayesian gene network generated from apoptosis timecourse data. Dots represents transcripts ("nodes") and arrows between the dots represent potential cause and effect interactions between transcripts ("edges"). A hypothetical master-regulator of apoptosis in these cells (GABARAP) is positioned at the top of the network graph (click image to see higher resolution version).


In the female reproductive system cyclical angiogenesis and vessel regression are precisely matched to the cyclical growth and regression of the ovary and endometrium. In Cambridge our group have investigated the role played by gene expression changes that may regulate angiogenesis in the endometrium and in the disease endometriosis (e.g. Human Reproduction; 10; 2356-2366, 2004, figure 5).


Figure 5. Scatterplot showing gene expression changes induced in human endothelial cells by soluble factors produced by proliferative-phase human endometrial epithelial cells (click image to see higher resolution version).


Inflammation involves the passage of white blood cells from the lumen of blood vessels into tissues, and appears to be controlled to a large extent by signalling cascades within endothelial cells. In a student project in Partnership with Pfizer, we used gene arrays to map the RNA transcript abundance changes that occur in endothelial cells during inflammation. This work has revealed inflammation-associated patterns of gene expression in endothelial cells and highlighted potential therapeutic targets. We are now using a new in vitro model of inflammation to test the validity of our individual gene array results. In this model we use RNAi techniques to "knock down" specific RNAs in endothelial cells, and use computerised video microscopy to quantify the effects of this treatment on the adhesion of flowing leukocytes to a stationary genetically modified endothelial cell monolayer. Dr Christopher Kirton from Cambridge University visited our lab in 2006 to help us set up this technique in Auckland. 


(2) Mechanisms of apoptosis of various cell types

Apoptosis is a genetically-programmed form of cell suicide. In addition to studying the RNA and cell surface carbohydrate changes that may contribute to this process in endothelial cells (described above), we have also studied the role of NFkB- and HIF-dependant transcript abundance changes in the regulation of neutrophil apoptosis (Walmsley et. al. Journal of Experimental Medicine; 201; 105-115, 2005, figure 6). A PhD student co-supervised by Cris has also used proteomics to study the mechanisms of action of Bcl-2 family apoptosis regulators. Cris has previously used gene knockouts to study the role of the apoptosis regulator Bcl-2 family members in male germ cells, gut and leukocytes (e.g. PNAS 96: 14943-8, 1999, PNAS 95: 12424-12431, 1998, Oncogene 19: 3955-3959, 2000).


Figure 6. Scatterplot showing gene expression changes induced in hypoxia in human neutrophils, Journal of Experimental Medicine; 201; 105-115, 2005 (click image to see higher resolution version).


(3) Research into new bioinformatic and gene regulatory network techniques

Gene array studies have revealed a great deal about cell biology and pathology. However, due to their technical limitations they have sometimes proved frustrating and have sometimes produced results that are very difficult to interpret. In collaboration with computational biologists, our group has worked on the best ways to use basic bioinformatic techniques to analyse gene array information (e.g. Schoenfeld Angiogenesis; 7; 143-156, 2004) and implemented novel methods such as Independent Component Analysis (ICA) (e.g. Saidi et. al. Oncogene; 23; 6677-6683, 2004, figure 7), as well as gene network methods (Bioinformatics; 24(7); 932-942, 2008; Tamada et al., Pacific Symposium on Biocomputing; in press, 2008) and Bayesian methods to combine in vivo and in vivo data (Bioinformatics; 23; 1936-44, 2007). These collaborations with computational biologists, who can use their expertise to drive new computational method development ,while we focus on biological interpretation and the laboratory components of the projects, have proved very productive.


Figure 7. Independant Component Analysis (ICA) reveals patterns of gene expression (called "components", denoted by *) associated with the loss of growth factor support and subsequent apoptosis in endothelial cells, taken from FASEB J 18; 188-190, 2003 (click image to see higher resolution version).


We have also combined siRNA gene knock-down techniques with Codelink' gene arrays to map the transcriptome changes that occur in endothelial cells when over 400 different signalling molecules and transcription factors are reduced in abundance. In collaboration with GNI ltd, we have used this gene array information to infer endothelial cell Bayesian gene regulatory networks to assist with target discovery. We are validating and tuning these network models using further siRNA experiments and combining them with time course gene array information about endothelial cell drug responses (e.g. Imoto et. al. Pacific Symposium on Biocomputing; 11; 559-571, 2006, figure 8). We believe that with careful use and extensive validation, gene array studies combined with Bayesian inference techniques provide valuable tools for target discovery and drug development (systems pharmacology). This is a fascinating cross-disciplinary area, which merges the fields of biology and mathematics, and translates research from academia to industry.


Figure 8. Conceptual view of our use of Bayesian gene network analysis for systems pharmacology, Pacific Symposium on Biocomputing; 11; 559-571, 2006 (click image to see higher resolution version).


In collaboration with scientists in Adelaide we have combined microarray information about mRNA and miRNA changes that occur in the disease endometriosis (Figure 9; Molecular Endocrinology; in press 2008).


Figure 9.

Like others, we have found that analysis at the level of pathways rather than at the level of individual genes is valuable for understanding the biology behind disease processes. We have taken this approach to understand how molecules in epithelial cells and the stromal cells that surround them interact in the disease endometriosis (American Journal of Pathology; 173 (3); 700-715, 2008) (click image to see higher resolution version).


(4) Gene Network analysis in cancer

In Auckland we are inferring gene regulatory networks in three types of cancer cell; melanoma (using A 375 cells), colon cancer, and breast cancer (using MCF-7 cells). We are inferring gene networks in these cells based on gene array analysis of siRNA experiments, in which approximately 70 transcripts are specifically "knocked down". This data is then combined with clinical microarray data sets form other groups. These project is run in close collaboration with the Auckland University Bioengineering Institute (Dr Edmund Crampin and Prof Peter Hunter), Dr Mik Black from the University of Otago, and the company GNI ltd. Ultimately, we hope to develop improved web-based tools that combine molecular and clinicopathological information to better predict individual patient prognosis and treatment outcome.

An example of a preliminary Boolean melanoma cell gene network based on a growth factor-deprivation timecourse experiment is shown in figure 10. This type of network is being combined with Bayesian gene networks and dynamical systems models using new computation methods we are developing as a framework that can be used by other groups.

We are also studying the function of the putative master regulators of cancer gene expression revealed by our gene networks using traditional cell biology and xenograft assays. We have a particular interest in discovering novel molecular signals that may mediate cross-talk between tumour cells and their supporting stromal cells such as endothelial cells and leukocytes.


Figure 10. (A) k-means clustering of A375 cell growth factor deprivation timecourse gene array data revels patterns of gene co-regulation as A375 melanoma cells leave the cell cycle and begin the process of apoptosis. (B) Boolean networks were inferred from this gene array data. Gene-to-gene connections are shown. The transcripts with the greatest connectivity to the regulation of other transcripts are placed nearest the top of the graph.


Selected 2007 to 2010 publications

Yamaji M, Bielby H, Licence D, Chen W, Cook E, Smith S, PRINT C and Charnock-Jones S. VEGF-A Loss in the Haematopoietic and Endothelial Lineage Exacerbates Age-Induced Renal Changes. Microvascular Research. In press 2010.

Mehta S., Shelling A., Muthukaruppan A., Lasham A., Blenkiron C., Laking G. and Print C. Predictive and Prognostic Molecular Markers for Cancer Medicine. Therapeutic Advances in Medical Oncology: 2, 125-148, 2010.

Tamada Y, Imoto S, Araki H, Nagasaki M, Print C, Charnock-Jones D S and Miyano S. Estimating Genome-Wide Gene Networks Using Nonparametric Bayesian Network Models on Massively Parallel Computers. IEEE/ACM Transactions on Computational Biology and Bioinformatics, In press 2010.

Jeffs A, Glover A, Slobbe L, He S, Woolley A, Print C, Baguley B, and Eccles M. A gene expression signature of invasion in metastatic melanoma cells. PLoS ONE; 24(12):e8461, 2009.

Hentschell D.,M., Harfouche R., Piecewicz1 S., Basu S., Print C, Eavarone D., Kiziltepe T., Sasisekharan R., Sengupta S. Regulation of vasculogenesis by glycosaminoglycans. Circulation; 120(19), 1883-92, 2009.

Ohlsson Teague E, Print C and Hull L. The role of microRNAs in endometriosis and associated conditions. Hum Reprod Update, 16(2):142-65, 2010.

Araki H, Tamada Y, Imoto S, Dunmore B, Sanders D, Humphrey S, Nagasaki M, Doi A, Nakanishi Y, Yasuda K, Tomiyasu Y, Tashiro K, Print C, Charnock-Jones S, Kuhara S and Miyano S. Analysis of PPARα-dependent and PPARα-independent transcript regulation following fenofibrate treatment of human endothelial cells. Angiogenesis; 12(3), 221-229, 2009.

Ohlsson-Teague M, Van der Hoek K-H, Van der Hoek M B, Perry N, Wagaarachchi P, Robertson S, Print C, and Hull L M. Differentially expressed microRNAs and their mRNA targets constitute molecular pathways associated with endometriosis. Molecular Endocrinology; 23(2); 265-75, 2009. (selected by Faculty of 1000 Medicine).

Brunet-Dunand S, Vouyovitch C, Araneda S, Pandey V, Vidal L, Print C, Mertani H, Lobie P and Perry J. Autocrine human growth hormone promotes tumour angiogenesis in mammary carcinoma. Endocrinology; 150;1341-1352, 2009.

Tamada Y, Araki H, Imoto, S, Nagasaki, M, Doi A, Nakanishi Y, Tomiyasu Y, Yasuda Y, Dunmore B, Sanders D, Humphries S, Print C, Charnock-Jones S, Tashiro K, Kuhara S, Miyano S. Unraveling dynamic activities of autocrine pathways that control drug-response transcriptome networks. Pacific Symposium on Biocomputing; 2009: 251-63, 2009.

Enis D, Dunmore B, Johnson N, Pober J and Print C. Anti-apoptotic activities of Bcl-2 correlate with vascular maturation and transcriptional modulation of human endothelial cells. Endothelium, 15(1); 59-71, 2008. (This paper gives examples of the use of pathway analysis and transcription factor target analysis with basic microarray data sets).

Hull M, Rangel C, Borthwick J, Doig J, Johnson C, Smith S, Tavaré S, Print C, and Charnock-Jones DS. Endometrial-peritoneal interactions during endometriotic lesion establishment. American Journal of Pathology; 173(3); 700-715, 2008. (This paper gives an example of the use of xenograft experiments to identify distinct molecular pathways in two different types of cell within the same tissue).

Hirose, O., Yoshida, R., Imoto, S., Yamaguchi, R., Higuchi, T., Charnock-Jones, D.S., Print, C., Miyano, S. Statistical inference of transcriptional module-based gene networks from time course gene expression profiles by using state space models. Bioinformatics; 24(7); 932-942, 2008.

Sykacek P, Clarkson R , Print C, Furlon R, Micklem G. Bayesian Modelling of Shared Gene Function. Bioinformatics; 23; 1936-44, 2007 Evans AL, Bryant J, Skepper J, Smith S, Print CG* and Charnock-Jones DS*. Vascular development in Embryoid Bodies. Quantification of transgenic intervention and antiangiogenic treatment. Angiogenesis 10; 217-226, 2007.

Cheng C-W, Bielby H, Licence D, Smith SK, Print CG, Charnock-Jones DS. Quantitative cellular and molecular analysis of the effect of progesterone withdrawal in a murine model of decidualisation. Biology of Reproduction; 76; 871-883, 2007.

Affara M, Dunmore D, Savoie C, Charnock-Jones S and Print C. Understanding Endothelial Cell Apoptosis: What can the transcriptome glycome and proteome reveal. Philosophical transactions of the Royal Society 362; 1469–1487, 2007. (This paper reviews general aspects of the biology and pathology of endothelial cell apoptosis, and the ways in which advanced bioinformatic methods can address questions in this field).

Duval H, Johnson N, Li J, Evans A, Chen S, Licence D, Skepper J, Charnock-Jones D S, Smith S, and Print C. Vascular development is disrupted by endothelial cell-specific expression of the anti-apoptotic protein Bcl-2. Angiogenesis 10; 55-68, 2007.