Att_6-_STO1_Analytic_capabilities.pdf

Page 1 of 3 Attachment 6 – Sample Task Order 1 Analytic Capability Assumptions October 1, 2024

Analytic Capability Assumptions: Sample Task Order 1

The Contractor should assume that analytic needs of the Division will have varying degrees of size and complexity as described in the Data Management assumptions above. Populations to be analyzed will range from small studies including a hand-full of subjects (e.g. family-based studies) to large pooling projects analyzing complicated covariate and laboratory data from tens of thousands of subjects spanning multiple time-points.

A few examples of analytic tasks to be supported under the contract include, but are not limited to, those listed below. This list is meant to be illustrative only and not comprehensive and support shall evolve to meet the changing analysis needs of the Division.

Routine statistical programming and analysis tasks

1. Create sample Statistical Analysis System (SAS) and STATA programs to extract data from analytic files.

2. Performing tests of significance including Chi-square, Student t and Wilcoxon t, Fisher’s exact, and Kolmogorov-Smirnov tests, ANOVA, ANCOVA which are routinely used for univariate analyses.

3. Utilizing techniques such as Linear and Logistic Generalized Linear Modeling (GLM) procedures to identify and control for the effects of confounding factors in multivariate analyses.

4. Performing correlation and regression methods for analyses of case-control studies (e.g., logistic regression, polytomous regression) and cohort studies (e.g., Poisson regression and Cox Proportional Hazards regression); tests for trends, interactions, and joint effects may be necessary.

5. Utilizing techniques for missing value imputations.

6. Analyzing data from repeated measurements for marginal and conditional analysis of longitudinal data.

7. Analyzing data from studies with complex design features, such as correlated observations, clustering, weighting, and stratification using specialized software (e.g., Survey Data Analysis, SUDAAN).

8. Calculating incidence, prevalence, mortality, odds ratios, relative risks, standardized mortality and incidence ratios (SMR/SIR), and temporal trends and survival (examples of survival analyses include Kaplan-Meier estimates and time-dependent Cox proportional hazards regressions).

9. Providing temporal trends in incidence and mortality rates and maps indicating geographical variation of incidence and mortality rates using regression methods; for this purpose, the Contractor shall utilize data resources including SEER, IARC, and others as deemed needed by DCEG investigators.

10. Providing graphical representations of epidemiology and laboratory assay data for effective visualization of trends and patterns. Such representations may be incorporated in formal reports or manuscripts. Examples of software used for graphical display within DCEG include but are not limited to, statistical packages in R, Excel, Origin, ArcGIS (ESRI, Redland, CA).

75N91024R00009

Page 2 of 3 Attachment 6 – Sample Task Order 1 Analytic Capability Assumptions

Molecular and Bioinformatic Analysis Tasks

1. For statistical/bioinformatics analyses of data resulting from molecular technologies which are essential to understanding gene-environment interactions that result in changes in cancer susceptibility among individuals.

2. For routine analyses including statistical tests for linkage disequilibrium (Hardy-Weinberg Equilibrium, HWE), and differential expression of genes, proteins, and metabolites.

3. To aggregate individual molecular data points to create derived covariates which may then be used in the epidemiological analyses to increase the strength of precision in risk factor prediction. Examples of such analyses include, but are not limited to, the integration of molecular covariates from gene expression/protein expression/metabolite profiling, pathway/network analyses, haplotype analyses, haploblock construction, and comparative genome analyses (e.g. phylogenetic trees) of human microbiomes to infer microbial composition. Models used to perform molecular epidemiology analysis shall take into account different modes of inheritance (recessive, dominant, additive), as well as epistasis, and population admixture determination. The Contractor shall provide support to perform these types of analyses and create streamlined analysis workflows for DCEG investigators.

4. To integrate biomarker data from molecular technologies with biological annotation from available bioinformatics resources in a meaningful and valid manner to facilitate interpretation of such data in the context of biological information.

5. Utilize cutting-edge statistical analysis methods to analyze high-dimensional data resulting from evolving molecular technologies used in DCEG’s studies. Meaningful and valid interpretation of high-dimensional molecular data often demands the use of computationally intensive analysis methods such as permutation tests, simulations, and data dimension reduction methods (for example, clustering, principal component analysis) and interactive visualization software to display these results graphically. The Contractor shall provide programming support for evolving computational methods involving complex mathematical and statistical modeling and simulations designed to analyze such data. The Contractor shall use innovative programming approaches to ensure scientific accuracy and computational feasibility, for example using parallel processing methodology via computing clusters (e.g., the Biowulf cluster at the NIH).

6. Perform confirmatory replicate study analyses of high-dimensional data from high-throughput molecular studies. The Contractors shall provide such support for novel experimental and analysis designs including meta-analyses (cross-study), and multi-stage replication studies and pooling data generated from multiple platforms to increase the power of detection.

Statistical Consultation and Methods Development

Previous examples of such support include:

1. Creating software programs and methods for cancer risk prediction and attributable risk calculations to analyze genetic and non-genetic contributions to cancer (http://dceg.cancer.gov/tools).

http://dceg.cancer.gov/tools

Page 3 of 3 Attachment 6 – Sample Task Order 1 Analytic Capability Assumptions

2. Creating software programs to study the behavior of methods and designs for genome-wide association studies such as multiphase designs to select promising SNPs for further study in later replication stages.

3. Creating software programs to compute sample sizes needed for innovative study designs or study objectives such as testing for gene-gene and gene-environment interactions.

4. Providing statistical support in developing new and innovative study designs, developing new methods of analysis such as detecting familial aggregation of disease from record linkage data.

5. Validating new cancer risk prediction models.

6. Analyzing data from next-generation sequencing.

7. Facile development of software for analysis methods developed by DCEG investigators (e.g., R packages and SAS Macros for computationally complex algorithms).

8. Capability to combine different software packages to implement an analysis or to develop a software tool (e.g., embed Fortran or C++ code into R).

9. Review analytical research plan provided by investigator(s), implementing existing methodology such as the risk of ovarian cancer algorithm (ROCA) and the joint modeling (JM) package in R.

10. Along with investigators, develop measures for evaluating the quality of the risk prediction and for comparing different predictors.

11. Adapt statistical approach to account for longitudinal biomarkers that are above and below a known detection limit. Further, adapt approach to account for missing data.

12. Write open source code for extending joint modeling approaches for multivariate longitudinal biomarkers.

13. Along with investigators, develop data analytic approaches for risk prediction from high-dimensional longitudinal biomarker data.

14. Develop graphical displays for presenting the relationship between longitudinal biomarkers and cancer risk.

15. Develop software for doing power calculations in this setting.

16. Produce flexible and efficient open source code that can be used by practitioners for risk prediction in this setting.

Tool and Software Development Assumptions

1. Any system or tool to be maintained or hosted on an internal (intranet) or external (internet) facing network shall have a minimum uptime of 99.5% per year.

2. If required, Contractor, is responsible for initially receiving and triaging (Tier 1) support requests from end users.

3. For ongoing operations and maintenance, all services that are in place shall continue and the parameters of the system will be maintained as described in the SOW.