"Large-scale gene screening, integration of multi-omics datasets, and network-based analysis can serve as valuable tools for understanding disease mechanisms, identifying disease dependencies, and discovering potential therapeutic targets for various diseases beyond cancer. However, applying similar methods to other diseases may be constrained by the specific biology and molecular mechanisms of each disease.Recently, researchers at the Wellcome Sanger Institute’s Cancer Genome Project, including Matthew J. Garnett, employed artificial intelligence (AI) methods to discover 370 new drug targets across 27 different types of cancers. The related paper was published online on January 11, 2024, in the journal ‘Cancer Cell.’
According to a press release from the Wellcome Sanger Institute, this research not only brings researchers closer to a comprehensive Cancer DepMap, which aims to identify specific cancer dependencies on a large scale for personalized treatment based on patients’ molecular features but also contributes to the development of targeted therapies for cancer.
Dr. Ying Pan, Head of the UK-Silicon Valley AI Hub, stated, “The greatest value of this research is that it may help formulate personalized treatment strategies based on the molecular characteristics of individual tumors. By identifying and prioritizing specific candidate targets, this research opens up new possibilities for the discovery of targeted therapies and anti-cancer drugs.”
According to a 2021 review in the journal ‘Nature,’ cancer “dependencies” are the genes, proteins, or other molecular features that tumor cells rely on during various processes such as growth and development. The Cancer Dependency Map project, initiated collaboratively by the Wellcome Sanger Institute, MIT, and the Broad Institute, aims to provide personalized treatment for patients.
In Garnett’s study, researchers integrated whole-genome CRISPR-Cas9 screening data for 27 types of cancers and 17,647 genes from the Sanger Project Score DepMap and Broad DepMap. They used AI methods to analyze the datasets deeply, identifying genes, proteins, and pathways crucial for the survival of cancer cells. Dr. Pan commented, “The newly introduced AI-based methods increased the utility of CRISPR-Cas9 screening on cancer cell lines by threefold, integrating various datasets, including clinically relevant transcription features, metabolic information, and proteomic data, for a more comprehensive analysis.”
Subsequently, Garnett and his team connected gene dependencies with clinical biomarkers using protein-protein interaction networks, ultimately generating a detailed list of candidate drug targets. Protein-protein interactions play a crucial role in various biological processes, significantly advancing drug development, disease treatment, and medical diagnostics.
The researchers wrote in the paper, “Almost all targets have a related biomarker, which is important because the genetic evidence between targets and diseases increases the likelihood of FDA (U.S. Food and Drug Administration) drug approval by 2-4 times.”
However, they also acknowledged in the paper that, in most cases, researchers do not develop drugs for these candidate targets. The presence of genomic-related biomarkers varies by cancer type, and some types of cancer lack genome-related targets, requiring new methods to expand the drug library.
Dr. Pan stated that the findings of this study are significant for cancer drug development. “This study identifies candidate tumor drug targets based on specific gene dependencies and prioritizes them, providing a potential avenue for developing targeted therapies. These findings can guide the discovery and design of new drugs or treatment strategies, selectively targeting discovered weaknesses in cancer cells. Integrating clinically relevant markers and screened cancer-driving genes enhances the potential clinical relevance of the identified targets.”
However, this study has limitations. Dr. Pan pointed out that the research primarily used immortal human cancer cell lines, which may not fully represent the complexity and specificity of patient tumors. Functional screening with cell lines may not capture all gene and molecular changes present in tumors. The study also acknowledged the need for further analysis in different cancer models, including underrepresented types, to deepen understanding of cancer dependencies.
He mentioned that while integrating various clinically relevant datasets is valuable, the study recognizes that not all available datasets have been included in the analysis of human cancer cell lines. For instance, as 23% of cancer-driving genes are excluded from cell lines, modeling dependency in specific cancer mutation backgrounds may introduce biases and limitations into the analysis.
Additionally, transferring the study results from cell lines to the clinical environment poses challenges. “Especially since the study did not include preclinical data in animals, which may limit the relevance and applicability of the study results in real-world clinical scenarios.”
“While the specific methods described in this study focus on cancer research, similar approaches can be applied to other diseases,” Dr. Pan said. Large-scale gene screening, integration of multi-omics datasets, and network-based analysis can be valuable tools for understanding disease mechanisms, identifying disease dependencies, and discovering potential therapeutic targets for various diseases.
Dr. Pan noted that applying similar methods to other diseases may be constrained by the specific biology and molecular mechanisms of each disease, as different diseases have unique characteristics and potential pathways of formation. Successful translation of research methods requires considering the use of different experimental models and clinical factors. Moreover, insights gained from cancer research may not necessarily generalize to non-oncological diseases with specificity. “Research techniques and methods may have some replicability, but the key is to consider the uniqueness of each disease to ensure the relevance and applicability of research results.”
Leave a comment