This author interview is by Dr Jie Zheng, of Nanyang Technological University, Singapore. Dr Zheng's full paper, In Silico Prediction of Synthetic Lethality by Meta-Analysis of Genetic Interactions, Functions, and Pathways in Yeast and Human Cancer, is available for download in Cancer Informatics.
Please summarise for readers the content of your article.
A pair of genes is called synthetic lethality (SL) if mutations of both genes will kill a cell while mutation of either gene alone will not. Hence, a gene in SL interactions with a cancer-specific mutated gene will be a promising drug target with anti-cancer selectivity. As wet-lab screening approaches are still expensive, computational methods are important for large-scale discovery of SL interactions.
In this article, we proposed a computational approach named MetaSL for predicting yeast SL, which integrates 17 genomic and proteomic features and the outputs of 10 classification methods. We also conducted analysis for feature ranking output by MetaSL, which provided biological insights into the observed SL in yeast. Moreover, through orthologous mapping from yeast to human genes, we then predicted several lists of candidate SL pairs in human cancer.
How did you come to be involved in your area of study?
When I was a postdoc researcher at USA National Institutes of Health, I worked on recombination hotspots in human genomes. After moving to Singapore as a junior faculty, I continued to work on this topic, but started to connect recombination directly with human diseases, especially cancer, for the critical roles of homologous recombination in genome instability and DNA damage repair. Through exploration of the literature, I came to know the new anti-cancer strategy of synthetic lethality. Based on this study, I applied for a grant on computational methods to understand and predict new SL gene pairs as potential cancer drug targets, and got awarded by the Ministry of Education, Singapore. This paper is part of this project.
What was previously known about the topic of your article?
There are several studies working on computational prediction for SL pairs. However, most of them focused on using individual features or single machine learning models. For example, Li et al. used protein domain as the main type of features and support vector machine (SVM) for SL prediction (Ref.  in the main text).
How has your work in this area advanced understanding of the topic?
Our work can shed light on both computational and biological aspects of SL prediction. Firstly, meta-analysis methods have been shown to improve the prediction accuracy by data integration and model combination. Secondly, we observed that similarity based features (e.g. relation between two proteins) are more important than lethality based features (e.g., characteristics for individual proteins). Thirdly, study of sig¬nalling pathways will be promising to understand and inter¬pret the underlying mechanisms of SL, as our manually collected SL pairs tend to occur in the same pathways.
What do you regard as being the most important aspect of the results reported in the article?
By comparing three types of human SL pairs (as shown in Table 6 in the main text of our article), our manually collected SL pairs are more likely to be involved in the same pathways than those predicted from yeast SL pairs. These results indicate that we may derive more features from signaling pathways for SL prediction. Our next step is to study the topological and dynamic properties of signaling pathways (e.g. crosstalk) to further improve the prediction of SL gene pairs.