A multi-layer encoder prediction model for individual sample specific gene combination effect (MLEC-iGeneCombo)

Using data from gene combination double knockout (CDKO) experiments, top ranked synthetic lethal (SL) gene pairs were highly inconsistent among different SL scores. This leads to a significant concern that SL prediction models highly depend on SL scores. In this paper, we introduce a new gene combination effect (GCE) measurement, log-fold change of dual-gRNA expression before and after CRISPR-cas9 lentivirus transfection. We show it is a direct and highly consistent measurement of GCE in all CDKO experiments. We therefore develop a multi-layer encoder model for individual sample specific GCE prediction, MLEC-iGeneCombo. Under a deep learning framework, MLEC-iGeneCombo is a systems biology model that contains sample specific multi-omics encoder, network encoder and cell-line encoder. For the first time, MLEC-iGeneCombo predicts GCE for a new cell. Using data from 18 CDKO experiments, MLEC-iGeneCombo achieves an average GCE prediction performance, 71.9%. All three encoders significantly improve the model’s prediction performance (p≤0.01), and their combined use yields the best GCE prediction performance. Our source code is available at https://github.com/karenyun/MLEC-iGeneCombo.Author summary: Understanding how pairs of genes interact is critical for deciphering the complexities of genetic diseases and cancer therapies. Our work focuses on accurately measuring how the combination of two genes affects cell survival, especially when both genes are disrupted. Traditionally, scientists have studied this phenomenon—called synthetic lethality—by evaluating indirect measures, leading to inconsistent and conflicting results. To address these inconsistencies, we adopted a direct and reliable measure in a gene combination double knockout experiment: log fold change of sgRNA expression. In our systematic review of existing experimental methods, we found that log fold change was the most consistent and comparable measure across various studies. Building upon this insight, we developed a new model capable of predicting gene combination effect in a new cell. Hence, its gene combination prediction is cell specific. Our work lays the foundation for personalized combinatory target discovery in cancer and other complex diseases.