*Equalized odds*, an important notion of algorithmic fairness, aims to ensure that sensitive variables, such as race and gender, do not unfairly influence the algorithm's prediction when conditioning on the true outcome. Despite rapid advancements, current research primarily focuses on equalized odds violations caused by a single sensitive attribute, leaving the challenge of simultaneously accounting for multiple attributes under-addressed. We bridge this gap by introducing an in-processing fairness-aware learning approach, FairICP, which integrates adversarial learning with a novel inverse conditional permutation scheme. FairICP offers a flexible and efficient scheme to promote equalized odds under fairness conditions described by complex and multi-dimensional sensitive attributes. The efficacy and adaptability of our method are demonstrated through both simulation studies and empirical analyses of real-world datasets.

SUMMARY: Identifying differentially expressed (DE) genes along cell pseudotime is crucial for understanding dynamic biological processes captured by single-cell RNA sequencing. However, existing DE methods either produce invalid P-values by ignoring the uncertainty in pseudotime inference or struggle to scale with the growing size of modern datasets. To address these limitations, we introduce PseudotimeDE-fast, a scalable method for detecting DE genes along pseudotime with well-calibrated P-values. Through comprehensive simulations and real-data analyses, we demonstrate that PseudotimeDE-fast delivers comparable or superior performance to existing approaches while offering substantial improvements in computational efficiency. AVAILABILITY AND IMPLEMENTATION: PseudotimeDE-fast is implemented in R with Rcpp acceleration and released under the MIT license. The source code is available at: https://github.com/dsong-lab/PseudotimeDE.