Table 1.
Machine- and deep-learning models for the prediction of synthetic lethality (SL).
Fig 1.
Overview of the proposed multi-layer encoder model, MLEC-iGeneCombo, for predicting gene combination effects.
The model comprises three components: a multi-omics encoder, a network encoder, and a cell-line encoder. The multi-omics encoder uses gene expression and essentiality data from the specific cell line. The network encoder incorporates gene population features into the PPI network as node features. The cell-line encoder processes population-level features of the cell lines as input.
Fig 2.
Correlation between gene essentiality and gene combination score.
Fig 3.
Correlation between gene expression and gene combination score.
Fig 4.
(A) shows results of omics feature selection for multi-omics encoder.
(B) shows average prediction performance of different submodules in 18 cell lines.
Table 2.
Prediction performance of our proposed multi-layer encoder model to predict gene combination effect for each cell-line.
Fig 5.
Testing results of separated sets C1, C2 and C3.
Fig 6.
This figure includes results of all ablation studies.
(A) shows results of input-layer ablation studies; (B) shows the results of hidden-layer ablation studies.