12 Match the unmatched

12.0.1 Introduction

The matching model has two layers. The first separates the sample into “the matched” and “the unmatched” group. The second layer further subsets samples from the “unmatched” that have similar tumor characteristic to “the matched”.

We performed Propensity Score Matching to select the “unmatched” patients from the patient dataset that best match to the “matched” group. Users can choose to perform this analysis based on Gender, Microsatellite status, and Tumor Mutational Burden (TMB). We use logistic regression to calculate propensity score with the optimal matching method.

12.0.2 How to start

Under the “Your Digital-twin community” tab, it includes three sub tabs: - The matched: Matched digital twin - The Unmatched: Match the unmatched samples with the matched samples so they have similar number and feature distribution. - Match and Unmatched: The summary total samples involved in matching.

12.0.3 Evaluate pre- and post- matching

This tab provides the main PSM analysis. It outputs “pre-matching statistics” and “post-matching statistics” to provide users with enough reference to decided whether to choose apply this model.

12.0.4 Review the matched and the unmatched

The matched and the unmatched are summarized together in an oncoplot and are saved for downstream analyses including survival comparison, tumor micro-environment and mutation gain/loss of functions predictions.