Three prizes ($3,000, $2,000 and $1,000) were awarded at the BioContact Québec symposium in October, 2006, where 12 finalists presented their research projects in poster form.
Recipients (2006)
Archived Recipients | 2005 |
BOUTROS Paul, University of Toronto, Toronto (October 2006)
A Modified Steepest Descent Algorithm for Enhanced Prognostic Marker Identification.
To help identify the molecular underpinnings of tumor behaviour, many groups have used expression arrays to identify sets of genes whose expression is correlated with patient prognosis. These gene sets serve two purposes. First, they provide insight into tumor biology, and may thus indicate therapeutic strategies. Second, they can serve as prognostic markers for tailored therapy, such as aggressive treatment for higher-risk patients. Despite some notable successes, prognostic markers identified in one patient cohort often perform poorly in other cohorts - even when sample sizes are large.
One of the major reasons for this inconsistency is the difficulty of selecting small numbers prognostic markers from amongst the thousands of genes on a microarray without over-fitting the data. The selection of prognostic subsets normally starts with a univariate analysis; this approach is based on the assumption that adding independently significant markers will create effective subsets. This assumption is not valid in most high-throughput datasets, possibly because of the contributions of multiple biological pathways. Further, traditional subset-selection algorithms are data-type dependent and computationally intensive.
To help circumvent this problem, a robust computationally efficient algorithm for prognostic marker identification, termed "modified steepest descent" was developed and validated.
A modified steepest descent algorithm for the identification of prognostic subsets was developed. The algorithm adds data points individually, and only retains those which maximize discriminative ability. This approach allows for data-type independence, and thus supports the analysis of heterogeneous datasets (e.g. concurrent analysis of genomic and proteomic datasets). Importantly, we have shown that modified steepest descent is highly efficient. For example, in a study of lung-prognostic markers, our method identified a prognostic subset in less than 100 iterations. The performance of this subset could not be significantly improved through months of exhaustive permutation-analysis, nor exceeded by any other subset selection method tested. Critically, the performance of this classifier has been validated in an independent patient cohort, confirming the reliability of modified steepest descent for prognostic marker identification.
This important new algorithm, termed "modified steepest descent", substantially out-performs other subset selection methods in both simulations and in a large-scale study of prognostic markers for non-small cell lung cancer.
TSUI Ivy F.L., University of British Columbia, Vancouver (October 2006)
High-Throughput Microarray CGH Detection of Genetic Alterations in Oral Pre-Malignant Lesions.
Oral cancer is the most common head and neck cancer accounting for 274,000 new cases in 2002 worldwide. Oral squamous cell carcinoma (OSCC), the most prevalent type of oral cancer, remains a significant health problem despite advances in surgery, radiation, and chemotherapy. The 5-year survival rate remains at 40% and is among the worst of all sites in the body and has not improved over the past 40 years. The real danger of this cancer is that in its early stages, it can be painless and patients are not aware of the presence of pre-malignant change. Thus, most cases are not diagnosed until the advanced stage of disease. If progressing pre-malignant lesions can be diagnosed and treated in the early stage, survival rates greatly improve to the 80-90% range.
The current criterion for clinicians to provide prevention and treatment strategy is based mainly on the presence and degree of dysplasia graded by pathologists. Oral pre-malignant lesions (OPL) are classified histologically into stages with increasing risk of developing into invasive SCC, namely from hyperplasia, mild, moderate, and severe dysplasias, and carcinoma in situ (CIS). High-grade pre-invasive lesions (severe dysplasias and CIS) have a high probability of progression into invasive carcinoma; in contrast, the majority of hyperplasia and mild/moderate dysplasias do not progress into cancer. Identifying causal genetic markers useful in the classification and prediction of behavior of early lesions will improve diagnosis and in turn prevent the disease.
The submegabase-resolution tiling set (SMRT) array comparative genomic hybridization (CGH) was employed to investigate genomic alterations of OPL in severe dysplasias and carcinoma in situ, pre-malignant disease with a high-risk of progression into invasive SCC. The whole genome has been arrayed as 26,819 bacterial artificial chromosome (BAC) derived amplified fragment pools spotted in duplicate (53,638 elements) resulting in tiling resolution with complete coverage of the sequenced human genome, and allowing the detection of breakpoints at a resolution of 0.08 Mbp. Sample DNA and normal reference DNA were differentially labeled with cyanine-dyes and competitively hybridized onto the array. Signals were subsequently scanned for each dye to determine the hybridization signal intensity ratios for each loci. DNA copy number profiles were generated by SeeGH custom software to display all data as log2ratio plots and visualize the chromosome profiles to facilitate the breakpoint detection of segmental DNA copy number changes across the entire genome.
Whole genome array CGH was used to generate high resolution segmental copy number profiles of 25 OPL graded as severe dysplasia and carcinoma in situ. Alignment of these profiles with the human genome map has resulted in fine-mapping of the numerous regional losses of 3p or regional gain in 3q and 8q. Furthermore, numerous novel segmental regions harboring candidate oncogenes and tumor suppressor genes were also identified in this study.
The generation of the high resolution segmental copy number profiles by SMRT array CGH allow us to detect novel segmental changes, to identify the oncogenes or tumor suppressor genes causal to carcinogenesis, and to precisely fine-map the breakpoints commonly known in OSCC development. Novel candidate oncogenes and tumor suppressor genes identified in this study will elucidate the molecular mechanisms underlying OSCC progression and may serve as genetic markers to help clinicians to manage OPLs.
ZHENG Jinzi, University of Toronto, Toronto (October 2006)
In Vivo Tracking of Liposomes Using CT and MR Imaging.
Currently, there are limited methods available for longitudinal and non-invasive in vivo assessment of the transport kinetics of carrier-based therapeutics, such as those relying on liposomes. Radioisotope labeling has been used to track small molecules in vivo with positron emission tomography (PET) and single photon computer tomography (SPECT)1-4. However, due to the short half-life of these labeling radioisotopes they are not suitable as tracers for longitudinal tracking of long-circulating carriers. Optical probes have also been explored as potential labels for carrier-based technologies5-7, but the limited tissue depth penetration and scattering issues associated with optical imaging systems make them inadequate for quantitative in vivo imaging applications. A viable strategy to non-invasively track the carrier in vivo is to employ the same carrier loaded with contrast-enhancing agents such as iodine and gadolinium. If these agents are retained within the carrier, they may act as surrogates for therapeutic molecules and allow for non-invasive tracking of the carrier in vivo using computed tomography (CT) and magnetic resonance (MR) imaging devices. If a correlation can be found between the change in signal intensities measured with CT and/or MR and the actual agent concentration detected in vivo, then this method has the potential to be successfully adopted for non-invasive quantification of pharmacokinetics and biodistribution studies. In this way, not only can the same animal be sampled over multiple time points, avoiding animal-to-animal variations, but the total number of animals required for each study can also be drastically reduced.
Specifically, this study evaluated the in vivo pharmacokinetics, in both mice and rabbits (Fig. 1), of a liposome-based CT and MR contrast agent previously developed by our research group8, 9 (Fig. 2). A dose of 6.3mg of iodine, 1.4mg of gadolinium and 7.0 mg of lipid was administered to each mouse via the tail vein. Mice (female Balb-C, 18-23g) were sacrificed at 12 different time points (5, 15, 30, 60 minutes, 6, 24, 48, 72 hours, 4, 5, 6, and 8 days following liposome injection, n=3 for each time point) and plasma samples were collected from each mouse via cardiac puncture. For the rabbit study, a dose of 63mg of iodine, 14mg of gadolinium and 70mg of lipid was administered to a female New Zealand White rabbit (2 kg). 1.5mL of blood was collected from the ear vein of the rabbit at each time point (5 minutes, 24, 48, 72 hours, 4, 5 and 7 days following liposome injection). Plasma samples were analyzed using high pressure liquid chromatography (HPLC, 245nm wavelength) to determine the iodine concentration, and inductively coupled plasma atomic emission spectrometry (ICP-AES) to measure the gadolinium concentration. Correlations were established between the iodine and gadolinium concentrations found in the rabbit plasma and the signal enhancement obtained in the rabbit aorta in CT and MR using circular regions of interest over all time points (Fig. 3). A linear relationship was found between the iodine concentration and Hounsfield Unit increase in CT, while an exponential relationship was observed between the gadolinium concentration and signal intensity increase in MR.
This pilot study demonstrated the potential of employing CT and MR imaging for longitudinal in vivo mapping of contrast agent containing liposomes.
