From Scientific Question to Genomic Analysis

I was approached to consult on the historical meaning of genomic variability in Yersinia pestis, the organism responsible for the Black Death, because of my background in genomics and infectious disease. The client wanted to build a scientifically rigorous project using historical Y. pestis genomic data, but had no defined research question, dataset, or analytical approach yet. I was working with a high school student with a broad but still introductory understanding of biology. The problem I solved was translating a real dataset generated with an advanced methodology, genomic sequencing and bioinformatics analysis, into a scientific project that he could understand, execute, and gain biological meaning from.

I served as the project’s manager, in effect, taking what looked like an unworkable goal and breaking it down until it became achievable. I narrowed his original interest into a specific, testable question: what is the relationship between year of genome collection and the number of nucleotide substitutions among Y. pestis 2.MED lineage samples from the Second Pandemic. I identified the publicly available Nextstrain Y. pestis dataset as the right source and showed him how to assemble a subset suited to that question, then worked through the study design with him, independent and dependent variables, inclusion criteria, and the null and alternative hypotheses. I built a sample R script and walked him through the statistical approach so he could run the analysis independently, then reviewed his work as it developed, not to check that the code ran, but to judge whether the conclusions he was drawing were substantiated by the data: whether he was mistaking a statistical association for a biological explanation, whether the pattern could be explained another way, and whether his conclusions stayed within what the dataset could actually support. That judgment, telling the difference between a result and a warranted conclusion, was the real work throughout the project, more than the coding, the statistics or the actual novelty of the findings.

Scientific consultation from framing through final review: guidance on the research question and study design, the dataset-selection strategy, a sample analytical script and statistical framework, and ongoing review of his developing analysis for whether its methods, results, and conclusions were logically consistent and properly bounded by the evidence.