Data Filtering for Genetic Perturbation Prediction (2503.14571v3)

Abstract: Genomic studies, including CRISPR-based PerturbSeq analyses, face a vast hypothesis space, while gene perturbations remain costly and time-consuming. Gene expression models based on graph neural networks are trained to predict the outcomes of gene perturbations to facilitate such experiments. Active learning methods are often employed to train these models due to the cost of the genomic experiments required to build the training set. However, poor model initialization in active learning can result in suboptimal early selections, wasting time and valuable resources. While typical active learning mitigates this issue over many iterations, the limited number of experimental cycles in genomic studies exacerbates the risk. To this end, we propose graph-based data filtering as an alternative. Unlike active learning, data filtering selects the gene perturbations before training, meaning it is free of bias due to random initialization and initial random selection. Moreover, reducing the iterations between the wet lab and the model provides several operational advantages resulting in significant acceleration. The proposed methods are motivated by theoretical studies of graph neural network generalization. The criteria are defined over the input graph and are optimized with submodular maximization. We compare them empirically to baselines and active learning methods that are state-of-the-art. The results demonstrate that graph-based data filtering achieves comparable accuracy while alleviating the aforementioned risks.