April 5, :45 AM 1:45 PM MARRIOTT MARQUIS HOTEL & MARINA, MIRAMAR VIGNETTE 1 VIGNETTE 2 VIGNETTE 3* VIGNETTE 4* VIGNETTE 5*

Transcription

1 April 5, :45 AM 1:45 PM MARRIOTT MARQUIS HOTEL & MARINA, MIRAMAR CHAIR: DANNY A. MILNER, JR., BRIGHAM & WOMEN S HOSPITAL, BOSTON, MA VIGNETTE 1 VIGNETTE 2 VIGNETTE 3* VIGNETTE 4* VIGNETTE 5* *VIGNETTES 3, 4, 5 WERE PREPARED BY SCOTT TOMLINS, MD, PHD, UNIVERSITY OF MICHIGAN, ANN ARBOR, MI ALL VIGNETTES WERE EDITED BY MARK E. SOBEL, MD, PHD, ASIP, BETHESDA, MD SPONSORED BY THE ASIP EDUCATION COMMITTEE

2 Vignette 1 A small molecule screening program using human endothelial cell lines that develop calcification in the presence of calcification-inducing media identifies a molecule that drastically reduces the process. Interestingly, this molecule has extremely high oral bioavailability and is eliminated by direct excretion in the urine. The working biological hypothesis is that the molecule enhances the activity of UBIAD1 (an intracellular cholesterol regulator). Your laboratory has a working model of the 5/6Nx rat chronic kidney disease system and a collaborator happens to have a CRISPR/CAS9 tool to replace the UBIAD1 gene with an inactive form of the protein. In your system, the 5/6Nx rats develop chronic kidney disease including vascular calcifications and you monitor the disease using a peripheral blood measure of creatinine prior to euthanasia. In your collaborator s system, serum calcium is elevated in the UBIAD1 inactive form in normal rats. You design an experiment to test the small molecule in your system as follows: At the end of 40 weeks, you measure the creatinine and calcium (see graph on the next page) of all of the rats and then sacrifice them. Using histology (H&E and Von Kassa stain) along with ImageJ (a free software program that allows you to do image based analyses, such as count cells or parse out a specific feature (nuclei, cytoplasm, etc) download from You quantify the amount of calcification in the kidneys and the heart (see graph on the next page).

3 Questions: 1. Are the differences in measurements of creatinine, calcium, and tissue calcification different between the groups? Which ones? 2. When approaching data such as this, a few questions need to be answered prior to beginning any analysis (and should best be thought of before designing the experiment!). These include the following: i. What kind of variables do I have? ii. What kind of statistical test(s) can I perform? iii. What kind of result am I looking for?

4 Vignette 2 A group of 1175 healthy subjects (43% Caucasian, 33% African or African American, 24% Hispanic/Latino) were recruited from college campuses in the Boston area (from among 26 different colleges) and were asked to provide a buccal swab for DNA sequencing along with a detailed questionnaire regarding their family history and medical health as well as a tube of blood for laboratory testing. They also agreed to complete a follow up survey every 5 years for the next 25 years in order to look at new diagnoses and diseases. For each patient, an aliquot of blood as well as the buccal swab were both used to sequence each patient to 40X coverage as well as perform comparative genomic hybridization to a sequenced and assembled reference genome. All genomes were cataloged for mutations included insertions/deletions, single nucleotide polymorphisms, and gene duplication. The survey included questions about all of the following: diabetes, hypertension, malignancy (specifically of breast, lung, colon, prostate, kidney and/or brain), infections (including frequency and specifically for mononucleosis, ear infections, head colds, urinary tract infections, toenail infections, persistent/excessive acne), diet, and exercise habits. All of the subjects were counseled to use a free pedometer (provided by the study team) which was connected to the internet and report their daily activity, which was monitored by the study. After 10 years (3 total surveys), a manuscript was published by a non-competing group in a mouse model showing that a specific mutation of pyruvate dehydrogenase kinase 4 (PDK4) caused a massive decrease in mouse activity as well as obesity in mice. You propose to look at the pedometer data of the study s subjects' activity to see if there is an association with fewer steps and mutations in PDK4. Your PI, however, thinks that such an association may be polygenetic (or even spurious in the mouse) and the entire genome should be examined in the context of all of the data. Questions: 1. How would you go about investigating any potential associations in your data set? 2. What statistical considerations are important in thinking about this question? 3. How should the pedometer data be parsed for the analysis?

5 Vignette 3 While investigating enzymes responsible for branched chain fatty acid metabolism in liver cells, you identify a novel protein that appears to interact with alpha-methylacyl-coa racemase (AMACR), an enzyme involved in fatty acid metabolism. Reading about AMACR, you discover that it is used diagnostically in prostate cancer pathology, as it is overexpressed in the majority of prostate cancers but not in benign prostate tissue. Realizing that your novel protein may also be overexpressed in prostate cancer, you hypothesize that expression of your novel protein may also have diagnostic utility in prostate cancer. Working with the genitourinary pathology pathologist at your institution, you obtain 20 transurethral resection of the prostate (TURP) specimens that were obtained after surgical intervention to relieve obstructive symptoms. The specimens all t contain benign prostate tissue. Likewise, you obtain 20 treatment-refractory prostate cancer metastasis samples obtained at biopsy or resection. After validating the performance of an antibody against your novel protein using positive and negative control cell lines, you perform immunohistochemistry for expression of your novel protein on the 40 tissue samples. You find moderate to strong staining in the majority of cancer cells in 18 of 20 prostate cancer metastasis samples, and weak staining in benign prostate epithelial cells in only one of 20 TURP specimens. Based on these results, you conclude that expression of your novel protein may have utility in the diagnosis of prostate cancer. One of the major areas to consider when evaluating the potential utility of a diagnostic biomarker is the area of intended clinical use. This often requires a deep understanding of the clinical issues regarding the diagnosis of the disease you are interested in, and will likely require consultation or collaboration with clinical experts. One of the most critical aspects to consider when planning the evaluation of a novel diagnostic biomarker is the current diagnostic process for the disease. For example, the majority of men in the US diagnosed with prostate cancer were identified from the population by an elevated level of serum PSA, which prompted a transrectal ultrasound-guided biopsy of the prostate, where multiple small cores of prostatic tissue are obtained with histopathologic evaluation by a pathologist. Most series find that about 30 to 40% of men undergoing prostate biopsy in the US are found to have prostate cancer on biopsy. In diagnostically challenging cases, the pathologist most commonly will utilize immunohistochemistry for basal cell markers (which are lost in prostate cancer), and AMACR (which is overexpressed in most prostate cancers). Prior to submitting your findings for publication and releasing a statement claiming that you have discovered a novel diagnostic biomarker for prostate cancer that you expect will be used in practice soon, what are the major issues to consider? 1) What are the sensitivity and specificity of your novel biomarker for the diagnosis of prostate cancer in the set of 40 tissue specimens? 2) How can you improve the cohort used to assess the diagnostic utility of your novel protein? 3) How can you evaluate whether assessment of your diagnostic biomarker is reproducible? 4) What experiments should you plan to best determine whether your biomarker will be adopted in clinical management? 5) If instead of developing a biomarker for prostate cancer, you instead found that your biomarker was a strong predictor of future development of a rare genetic disease, how would that impact your assessment of the potential diagnostic utility?

6 Vignette 4 While investigating mirna expression in the blood of mouse models of breast cancer, you identify the ratio of mir-101/mir-25 as being highly associated with development of metastatic disease. You hypothesize that the ratio of these markers in human blood may be prognostic for patients with breast cancer. Your institution has an archive of blood samples taken from patients prior to lumpectomy for breast cancer. You are able to obtain 15 samples from patients who did not have disease recurrence within 5 years after lumpectomy, and 15 samples from patients whose disease recurred within 5 years after lumpectomy. You develop real-time RT-PCR based assays compatible with the archived blood samples that are capable of assessing mir-101 and mir-25. You successfully assess all 30 samples and calculate mir-101/mir-25 (mir ratio) for the samples as shown in the Table below. You run into your mentor on the way out the door for the day and arrange to go over the results the next morning. That night, you research predictors of recurrence in breast cancer, and discover that a number of pathological parameters at the time of lumpectomy, such as tumor grade and size of tumor, are associated with the risk of recurrence. Likewise several other blood based biomarkers have been assessed for their ability to predict recurrence, including cancer antigen 15-3 (CA 15-3), which you know was run on the blood samples you assessed. No Recurrence Recurrence mir ratio CA 15-3 mir ratio CA 15-3 (ng/ml) ) What statistical tests can you use to determine whether the mir ratio is associated with the risk of recurrence in your data set? 2) Your mentor can access a cohort of 200 samples and wants to know what is the best mir ratio value (cut off) to decide whether a sample is positive or negative. How can you assess the diagnostic performance across the range of potential mir ratio values (0-100)? 3) What tests can you use to determine whether the mir ratio is a better biomarker for recurrence than (CA 15-3)? 4) Is it important to consider the pathological parameters in your evaluation of the potential clinical performance of the mir ratio? If you were to design a large validation study, how would you try to account for the clinical parameters, and how can you determine whether the mir ratio may have clinical utility?

7 Vignette 5 In order to understand the toxicity of rattlesnake venom, you are exposing human endothelial cells to vehicle or rattlesnake venom, and performing gene expression profiling using microarrays that assess the expression of 20,000 genes to identify those that are differentially expressed. Across triplicate experiments, you identify a set of 200 genes that are overexpressed greater than 2 fold (on average) in cells exposed to venom vs. vehicle, and a set of 50 genes that are under-expressed greater than 2 fold (on average). In a previous experiment, you exposed human endothelial cells to Salmonella, which identified a set of 1000 overexpressed genes, 7 of which were also in the set of 200 overexpressed genes in the rattlesnake venom experiment. Examining the list of under-expressed genes, you recognize several that are associated with protein ubiquitination based on your reading of the literature. Using Gene Ontology ( you identify a curated set of 35 human genes associated with protein ubiquitination (GO: ), 3 of which were in the set of 50 under-expressed genes in your rattlesnake venom experiment. 1) You wish to know whether there is more overlap in the overexpressed gene sets from the rattlesnake venom and Salmonella microarray experiments than you would expect by chance. What statistical approaches can you use to perform this assessment? 2) Likewise, what additional information do you need to know about the gene ontology database to determine whether there is more overlap in the set of genes under-expressed in rattlesnake venom and the GO protein ubiquitination gene set? 3) You find a database of 10,000 microarray experiments performed on the same platform as yours. What considerations must be taken into account if you are going to compare your differentially expressed gene sets to differentially expressed gene sets from this database of 10,000 experiments?