Fig. 1

Evaluation strategy of methods designed for transcriptome analysis in paired single subject samples. Motivation: Identifying the gene products altered between two conditions in a single subject without replicates (ss-DEGs) is highly relevant in precision medicine. While conventional analytical methods may be applied to discover differences between isogenic replicates studied in distinct conditions (r-DEGs), precision medicine has helped usher in the possibility that diagnosis, prognosis, and therapeutic choices may be determined more accurately from single-subject measurements. Accurate ss-DEGs methods enable studying (i) cancer vs unaffected adjacent tissue or (ii) an ex-vivo cellular provocation assay operating on relevant tissue with or without therapy. Evaluation framework. Step 1. A dataset comprising multiple biological replicates of isogenic transcriptomes observed on samples taken in distinct biological conditions is identified. Step 2 The replicates are split into two groups of independent samples: a reference set and a single-subject (ss) prediction set. Step 3. Each r-DEG method (e.g., EdgeR, DESeq, etc.) is applied independently to the reference set to generate multiple reference standards, as each method has biases and none can be truly considered as a gold standard (Step 3, top panel). The reference set consists of biological replicates between two conditions of isogenic samples, and is thus a proxy for studying and mimicking the isogenic biologic variation of one subject (and each set of r-DEGs is an attempt at becoming a gold standard. In parallel, each ss-DEG method is applied to independent pairs of samples (one in each condition) taken from the prediction set, each as a proxy to a single subject (Step 3. Bottom panel). Step 4. Accuracy scores are determined for each ss-DEG method against each r-DEG-derived reference standard. Step 5. Summary statistics are conducted across all experiments to determine the best ss-DEGs according to the conditions of application