Statistical Consultation, Statistical Methods Development, and Mathematical Modeling for Immunological Assay and Immunoassay
Tyson H. Holmes, Ph.D.
Statistical Director, Human Immune Monitoring Center
Institute for Immunity, Transplantation and Infection
Stanford University School of Medicine
“We are here to help.”
The purpose of this service is very specific and two-fold.
· To provide investigators with guidance on how to analyze their data themselves, including instruction on applicable R and/or SAS script.
· To develop new statistical methods or other mathematical models where existing methods will not permit adequate analysis.
Selected Assay Technologies:
· Luminex® platform for secreted soluble proteins
· Flow and mass cytometry, including intracellular staining
Selected Current Research Interests:
· Nonspecific binding, batch effects, antigen-excess, and other assay artifacts
· Optimization of immunological analysis pipelines
· Cluster analysis and gap finding for immunological and immunoassay data
· Estimation of and computations on biological networks (graph theoretic)
· Low-dimensional visualization and testing for high-dimensional data
· In vivo standard (calibration) curve development
· Microbiome characterization
· Distribution-free, regularized regression modeling of immunological and immunoassay data
Selected Survey of Publications:
· Holmes TH. 2020. Generalized mathematical model for immunoassay interference. Autoimmunity Reviews 19:102663.
- · Holmes TH, Subrahmanyam PB, Wang W, Maecker HT. 2019. Penalized supervised star plots: Example application in influenza-specific CD4+ T Cells. Viral Immunology 32:102-109.
· Holmes TH, He X. 2016. Human immunophenotyping via low-variance, low-bias, interpretive regression modeling of small, wide data sets: application to aging and immune response to influenza vaccination. Journal of Immunological Methods 437:1-12.
· Holmes TH, Lewis DB. 2014. Bayesian immunological model development from the literature: example investigation of recent thymic emigrants. Journal of Immunological Methods 414: 32-50.
Many consultations will not require data sharing. For those rare occasions that do, please proceed as follows.
- Data must be fully de-identified of...
- Protected Health Information (PHI) (https://acp.stanford.edu/hipaa/hipaa-faq) and
- any other high-risk data (https://uit.stanford.edu/guide/riskclassifications).
- Do not provide us with any information that would allow us to re-identify the data.
- Provide concise description of the dataset, its variables, and any limitations in the data, including missing values, low microbead counts, measurements above or below limits of detection, serial dilutions, etc. Here too, do not send any PHI (https://acp.stanford.edu/hipaa/hipaa-faq) or any other high-risk data (https://uit.stanford.edu/guide/riskclassifications) and do not provide us with any information that would allow us to re-identify the data.
- Studies must already have all applicable and necessary Institutional Review Board review and approvals, including for sharing data with us (https://researchcompliance.stanford.edu/panels/hs).
- Structure data with columns as variables and rows as observations. Longitudinal data will require multiple rows per participant.
- First row, and first row only, of data set will always contain variable names. Variable names should not include any special characters, including spaces or hyphens. Each variable name should begin with a letter of the alphabet.
- Contact Dr. Holmes (email@example.com) for acceptable file formats.
- * Contact Dr. Holmes regarding how to submit your data.
For more information on this data policy, please contact the HIMC Director,
Dr. Holden Maecker (firstname.lastname@example.org).
· Please do not begin by sending any data. Instead, please send via encrypted email to Dr. Holmes (email@example.com), copying Dr. Maecker (firstname.lastname@example.org), a one-page summary of the study design, primary hypotheses, and the type(s) of data collected.
· This one-page summary cannot include any PHI (https://acp.stanford.edu/hipaa/hipaa-faq) or other high-risk information (https://uit.stanford.edu/guide/riskclassifications) or any access to any PHI or other high-risk information.
· Dr. Holmes will respond promptly with an estimate of “hours not to exceed” required for the work.
Existing Tools for Download:
Please direct all questions regarding any of the utilities posted here to Tyson Holmes (email@example.com).
R utility for correcting for plate/batch/lot and nonspecific binding artifacts (Updated 8 Apr 2021). April 8, 2021: The R script has been updated to add the dpMFI values back to the mean pMFI values per SP in the output file (not in Figure 4). Use this version with revised output file to make comparisons among cytokines more meaningful in downstream analyses.To avoid a harmless warning message, download and install version 1.5.2-1 of emmeans. We thank Heather Pankow (Department of Psychiatry and Behavioral Sciences, Stanford University) for helping us to identify an error in the R script.
R utility is here for producing penalized supervised star plots. The paper that introduced and details penalized supervised star plots is published in Viral Immunology 32(2):1–8.