http://health.usnews.com/articles/health/cancer/2008/10/23/lung-cancer-gene-discovery-a-sign-of-cancers-future.html

These are excerpts from http://www.ostp.gov/galleries/PCAST/pcast_report_v2.pdf.

• “The Federal government should make critical investments in the enabling tools and resources essential to moving beyond genomic discoveries to personalized medicineproducts and services of patient and public benefit.”
• “FDA should implement a more transparent, systematic, and iterative approach to the regulation of genomics-based molecular diagnostics.”

“Molecular diagnostics can be used in a variety of ways to inform personalized medicine

• Assess the likely efficacy of specific therapeutic agents in specific patients.
• Identify patients who may suffer disproportionately severe adverse effects from a given treatment
  or dosage.
• Determine optimal dosages for drugs whose therapeutic effect is known to vary widely.
• Assess the extent or progression of disease – Molecular diagnostics have the potential to provide more accurate and timely information on disease prognosis or treatment effectiveness than the imaging and pathology methods currently used for this purpose, though future diagnostic approaches may integrate all of these methods.
• Examine surrogate measures for clinical outcomes – Researchers are investigating whether  biomarker-based molecular diagnostics can provide reliable proxies for longterm outcomes such as relapse or survival. Such tests could be used to shorten the length and expense of clinical trials.
• Identify patients who can benefit from specific preventive measures

Clinical Decision Support

• To date, few genomics-based diagnostic tests have reached the market, and these few products have been targeted primarily at clinical specialists and subspecialists who have been able to assimilate them into practice without special measures. However, if the number of innovative personalized medicine diagnostics and linked diagnostic therapeutic combinations reaching the market increases substantially, widespread adoption of these products

Molecular Diagnostics

• The ability to generate genetic profiles using microarrays and sequence-based approaches promises to expand greatly the utility of genetic tests in clinical medicine. This is because human diseases resulting from a single genetic alteration are rare. Most common diseases including cancer, cardiovascular disease, and diabetes result from a variety of genetic changes acting in concert. Moreover, the exact combination of genetic factors resulting in a specific disease often varies among individuals. To address this complexity, many companies and academic groups are developing complex molecular diagnostics (including IVDMIA tests based on microarrays) with the goal of establishing correlations between a specific pattern of genetic modification and/or gene expression and disease outcomes such as progression, response to therapy, or adverse reactions. In some cases, these correlations and their predictive value are strong enough that the tests can have clinical utility even in the absence of a full understanding of the effects of and interactions among the component genes. As with genome-wide association studies, there are many pitfalls in establishing robust and reliable disease correlations for genomic profiling tests.46, 47 Reproducible sample collection and processing is essential to avoid artifacts in gene expression patterns due to cell population subtypes or effects of processing on the apparent levels of expression. Standards for the measurement, analysis, and reporting of biomarker data are essential to allow data to be compared across different studies and different laboratories and reduce duplication in defining assay methods and data requirements. Complicated statistical methodologies are required because probing the expression of 10,000 or more genes can lead to spurious correlations simply by chance. Moreover, these studies require not only large sample sizes but also validation using independent sample sets.

The following statement issued by the office of the President of the United States answers this question in the following summary (see http://www.ostp.gov/galleries/PCAST/pcast_report_v2.pdf):

“‘Personalized medicine’ refers to the tailoring of medical treatment to the individual characteristics of each
patient. It does not literally mean the creation of drugs or medical devices that are unique to a patient, but rather the ability to classify individuals into subpopulations that differ in their susceptibility to a particular disease or their response to a specific treatment. Preventive or therapeutic interventions can then be concentrated on those who will benefit, sparing expense and side effects for those who will not.

“The principle of adjusting treatment to specific patient characteristics has, of course, always been the goal of physicians. However, recent rapid advances in genomics and molecular biology are beginning to reveal a large number of possible new, genome-related, molecular markers for the presence of disease, susceptibility to disease, or differential response to treatment. Such markers can serve as the basis of new genomics-based diagnostic tests for identifying and/or confirming disease, assessing an individual’s risk of disease, identifying patients who will benefit from particular interventions, or tailoring dosing regimens to individual variations in metabolic response. These new diagnostics can also pave the way for development of new therapeutics specifically targeted at the physiological consequences of the genetic defect(s) associated with a patient’s disease.

“The current high level of interest in personalized medicine from a policy perspective is attributable not only to the promise of improved patient care and disease prevention, but also to the potential for personalized medicine to positively impact two other important trends – the increasing cost of health care and the decreasing rate of new medical product development. The ability to distinguish in advance those patients who will benefit from a given treatment and those who are likely to suffer important adverse effects could result in meaningful cost savings for the overall health care system. Moreover, the ability to stratify patients by disease susceptibility or likely response to treatment could also reduce the size, duration, and cost of clinical trials, thus facilitating the development of new treatments, diagnostics, and prevention strategies.”

“Correspondence analysis provides tools for analyzing the associations between rows and columns of contingency tables. A contingency table is a two-entry frequency table where the joint frequencies of two qualitative variables are reported. For instance a (2 x 2) table could be formed by observing from a sample of n individuals two qualitative variables: the individual’s sex and whether the individual smokes. The table reports the observed joint frequencies. In general (n x p) tables may be considered.

“The main idea of correspondence analysis is to develop simple indices that will show the relations between the row and the column categories. These indices will tell us simultaneously which column categories have more weight in a row category and vice-versa. Correspondence analysis is also related to the issue of reducing the dimension of the table…” (p. 341)

Applied Multivariate Statistical Analysis

“A frequently applied paradigm in analyzing data from multivariate observations is to model the relevant information (represented in a multivariate variable X) as coming from a limited number of latent factors. In a survey on household consumption, for example, the consumption levels, X, of p different goods during one month could be observed. The variations and covariations of the p components of X throughut the survey might in fact be explained by two or three main social behavioral factors of the household…These unobserved factors are much more interesting to the social scientist than the observed quantitative measures (X) themselves, because they give a better understanding of the behavior of households.

“How can we provide a statistical model addressing these issues and how can we interpret the obtained model? This is the aim of factor analysis.” (p. 276)

Applied Multivariate Statistical Analysis

“…Two partners in a human chromosome pair…often line up next to one another in parallel array, look each other over, compare their respective DNA sequences, and then swap genetic information. One frequent result is that a gene sequence present on one chromosome will now replace the corresponding sequence carried by its partner. Before this information transfer, two distinct versions of a gene may have resided on the two paired chromosomes; afterward, one of these versions is lost, being replaced by a duplicated version of the gene originally present on the other chromosome. The result is two identical copies of a gene in a cell that previously carried two dissimilar versions.” (p. 75)

“Estrogen is a fully natural hormone, native to the body, yet it contributes to carcinogenesis in the breast and ovaries. In the breast, it drives the proliferation of cells lining the milk ducts during the menstrual cycle and pregnancy. The monthly multiplication of these mammary epithelial cells is followed by their die-off, this cycle repeating itself over and over in most women from menarche to menopause–generally between ages twelve and fifty.

“Many researchers trace the roots of breast cancer to these repeated bouts of estrogen-driven proliferation. The increased incidence of this disease in modern times seems connected to dramatically increased menstrual cycling. Due to greatly improved nutrition, menarche begins four or five years earlier in late-twentieth-century girls than it did in their great-grandmothers. In addition, reproductive practices have changed in Western society. Childbearing and breast feeding, both of which suppress menstrual cycling, are now postponed and, when they occur, encompass only a few years of adult life, unlike a century ago, when three decades of a woman’s life were often involved in cycles of birth and lactation.” (p. 60-61)

From Hofmann W (editor). 2006. Gene Expression Profiling by Microarrays, Clinical Implications. Cambridge University Press.

“Approximately 5-10% of all breast cancers are of hereditary origin, and two major breast cancer susceptibility genes have been identified to data, BRCA1 [19] and BRCA2 [20]. Although these high-penetrance syndromes account for a small proportion of all cancers, the identification of these genes and the investigation of their roles in breast cancer development and progression emphasize the genetic aspect of cancer in general, and provide a basis for extrapolating findings from ‘genetically defined’ studies to the investigation of the more common forms of sporadic disease.

“While mutation screening in the two known breast cancer susceptibility genes for hereditary breast cancer families, allowing mutation carriers to make informed decisions regarding surveillance and/or prophylactic approaches, has become commonplace at oncogenetic clinics across the world, the techniques used for screening are time-consuming and expensive. Studies of the histopathological features, genomic alterations and hormone receptor levels in these tumors support the notion that breast cancers caused by germline mutations in BRCA1 and BRCA2 differ from each other and from tumors not caused by mutations in these genes at a molecular level. While certain characteristics, such as medullary histology and ER negativity, are more common among BRCA1 derived breast tumors, which make up a somwhat homogeneous group, BRCA2 derived breast tumors and, to an even greater extent, non-BRCA1/2 (BRCAx) breast tumors constitute considerably more heterogeneous groups. Therefore, an alternative means of classifying BRCA1, BRCA2 and non-BRCA1/2 associated tumors would greatly facilitate the identification of patients carrying mutations in these genes. Moreover, comprehensive understanding of the underlying defects causing the development of hereditary breast cancers may greatly improve both treatment strategies and intervention options for the affected patients, and may give insights into breast cancer biology in general.” (p. 136-137)

“While Wang and coauthors, in concordance with van’t Veer and colleaugues, suggest an improvement in the risk assessment of breast cancer patients with the use of gene expression profiles compared to conventional criteria commonly used in Europe (St. Gallen, [42]) and the USA (NIH, [43]), the overlap in genes involved in the two predictors is minimal. Differences in techniques used and, maybe more importantly, patients included may account for a certain degree of these discrepancies; this demonstrates the need for transparency in terms of already published results as well as the implementation of well-designed, large, prospective studies aimed at answering very specific questions in terms of outcome prediction. While gene expression-based techniques lend promise to the improvement of individual patient care, these studies also illustrate the heterogeneity of the disease and that we have not yet reached a point where these applications can be readily incorporated into clinical practice.” (p. 149)

“Response to therapy: As mentioned, although ER status is predictive of response to hormonal treatment, the predictive value is not 100%. Moreover, there are, to date, no clinically useful predictive markers for a patient’s response to chemotherapy. In addition, all patients eligible for chemotherapy receive the same treatment although the average expected benefit is low; some patients fail to respond and suffer unnecessary toxicity, while others would benefit from a more severe treatment. Hence, selection of patients most likely to benefit from adjuvant systemic therapy would be a great advance in the clinical management of breast cancer. The assessment of one or a few individual markers has not been shown to be powerful enough to reveal the complex biology of clinical breast cancer and response to therapy. However, patterns of expression of larger numbers of genes could be successful in distinguishing between sensitive and resistant tumors.” (p. 149)

“As it has been illustrated by a few research groups, probably the most plausible approach for translating microarray-derived gene expression profiles into clinically applicabile routine assays is first to identify diagnostic or prognostic gene expression profiles consisting of a small number of genes using whole genome microarras (and fresh frozen tissues), and then validating the clinical efficacy of these genes in prospective studies using a simple and robust conventional assay such as RT-PCR and formalin-fixed paraffin-embedded tissue. Such an assay could be incorporated readily into clinical practice, and would also be the most cost-effective. While this strategy may improve and refine the stratification of patients greatly into available treatment alternatives, the greatest obstacle in the advancement of clinical oncology is the identification and development of novel, targeted drugs with optimal efficiency.” (p. 157)

“Microarray mRNA expression analysis is a powerful screening technique that tests a sample of 1 ugram total RNA with many thousands of parallel hybridizations. Even if each of these hybridizations is performed with precision and high reproducibility [48], the measures required to protect against false positives (random fluctuations exceeding the permitted threshold of variation) are draconic. For instance, on an array configuration with 45,000 probe sets and 20,000 expressed transcripts, the preset P-value used in differential expression analysis has to be P<10^-4 in order to have two or fewer false positives. Since microarray experiments are expensive, investigators avoid high numbers of repetition, often three times for well-defined cell preparations [48]. This means that, by accepting P<10^-4 in order to prevent false positives, one will lose most true positives as well, a situation that is not desired. We have learned from a number of related microarray experiments on the same biological object that linking the different data sets is a powerful way of guiding the investigator towards appropriate gene prioritization.

“Suppose beta-cell mRNA expression is measured in conditions A vs. B and we want to know which of the 20,000 transcripts called present are truly changed. When the frequency of true positives is 1% and the acceptance threshold P-value of each compared transcript is 1% and the acceptance threshold P-value of each compared transcript is 0.01, the differential expression analysis A vs. B would generate a data set of 200 true and 200 false positives. In terms of follow-up, this is a large project, as the obvious way to find out which are the true and which are false positives is to validate the transcripts one by one with other techniques like real-time RT-PCR. However, such analysis is an enormous task that would fill the time alloted to several Ph.D. students. But suppose that we test beta-cell mRNA expression a second time, now in conditions C vs. D. With a frequency of true positives of 1.5% and a threshold P-value of 0.01, we would now generate a second data set with 300 true and 200 false positives. When we accept as null hypothesis that the two comparisons bear no other than a random relationship in terms of behavior of RNA molecules, then the expected number of true and false positives exhibiting differential expression in A vs. B and C vs. D would be between 4 and 5.” (p. 203)

There is a page that explains how to do this, so I won’t repeat it, but here is a link: http://oct2007.archive.ensembl.org/info/software/website/installation/ensembl-data.html

This Wired article talks about the NetFlix data mining competition. They have offered a $1 million prize to whomever can beat their movie-recommendation algorithm by 10%. That seems like it should be completely feasible. Yet complex math hasn’t yet been the solution. They have reached 8+% improvements, but so far nobody has cracked that 10% barrier and claimed the prize.

Common sense tells me it should not be so hard to attain such an improvement, but of course that’s easy for me to say…

A “psychologist” has done quite well so far by taking into account “human factors” in addition to math. The story is dramatized for effect, but it illustrates the value in thinking about the context when trying to solve difficult quantitative problems. Otherwise, you may be shooting too much in the dark, and the high dimensionality of the data gets in the way. No matter how complex your model, pure math won’t always cut it; and as the author suggests, you may be prone to overfitting the model.

There are parallels in mining genetic data sets.  There have to be betters ways to look at these data sets and take the biological context into account. Everyone is excited about pathway analysis, and I can see some logic in that. But I think that’s only the beginning. I’d tell you my other ideas, but that might give away my strategic advantage in getting published in Nature or Science.