Constance H. Li
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Constance H. Li has a total of 50 co-authors. Here are the top co-authors by number of shared publications:
- Paul C. Boutros
Department of Human Genetics, University of California Los Angeles
10 shared publications - Robert G. Bristow
Manchester Cancer Research Centre and Cancer Research UK Manchester Institute, The University of Manchester, Manchester, UK
3 shared publications - Stephenie D. Prokopec
Princess Margaret Cancer Centre, University Health Network, Toronto, Ontario, Canada
3 shared publications - Vinayak Bhandari
Department of Medical Biophysics, University of Toronto, Toronto, Canada
3 shared publications - Chenghao Zhu
Department of Human Genetics, University of California, Los Angeles, CA, USA
2 shared publications - Jieun Oh
University of California, Los Angeles 1 Jonsson Comprehensive Cancer Center Los Angeles, California
2 shared publications
Latest Publications
- Open Access
Integrative single cell analysis of CD8+ T-cells across early and advanced oral cancers reveals signatures of anti-tumour activity
Tumour-targeting CD8 ⁺ T cells drive responses to every major form of cancer immunotherapy. Identifying them, however, remains an unsolved problem in solid tumours. The antigens they recognize are rarely defined and almost never shared between patients. We profiled 51,459 CD8+ T cells by paired single-cell RNA and T-cell receptor sequencing across 28 samples from 17 HPV-negative oral cancers spanning primary tumours, draining lymph nodes, metastases, and pembrolizumab-treated recurrences. We found that clonotypes that were expanded and shared across anatomical sites and timepoints were enriched within tumours and progressively selected over disease evolution and checkpoint blockade. Designating these shared-expanded clones as putative tumour-targeting cells, we trained a machine learning classifier that identifies them from transcriptome data alone. This 108-feature random forest signature recapitulated programmes of tumour reactivity and generalized to an integrated atlas of 89,318 CD8+ T cells from independent cohorts, showing progressive enrichment from normal to malignant tissue, and localized to tumour-proximal niches in spatial transcriptomics. By demonstrating that clonal behaviour across space and time encodes tumour reactivity in the transcriptome, this work establishes a generalizable framework for mapping tumour-engaged immunity without knowledge of the underlying antigen.
Open AccessIntegrative single cell analysis of CD8+ T-cells across early and advanced oral cancers reveals signatures of anti-tumour activity
Tumour-targeting CD8 ⁺ T cells drive responses to every major form of cancer immunotherapy. Identifying them, however, remains an unsolved problem in solid tumours. The antigens they recognize are rarely defined and almost never shared between patients. We profiled 51,459 CD8+ T cells by paired single-cell RNA and T-cell receptor sequencing across 28 samples from 17 HPV-negative oral cancers spanning primary tumours, draining lymph nodes, metastases, and pembrolizumab-treated recurrences. We found that clonotypes that were expanded and shared across anatomical sites and timepoints were enriched within tumours and progressively selected over disease evolution and checkpoint blockade. Designating these shared-expanded clones as putative tumour-targeting cells, we trained a machine learning classifier that identifies them from transcriptome data alone. This 108-feature random forest signature recapitulated programmes of tumour reactivity and generalized to an integrated atlas of 89,318 CD8+ T cells from independent cohorts, showing progressive enrichment from normal to malignant tissue, and localized to tumour-proximal niches in spatial transcriptomics. By demonstrating that clonal behaviour across space and time encodes tumour reactivity in the transcriptome, this work establishes a generalizable framework for mapping tumour-engaged immunity without knowledge of the underlying antigen.
- Open Access
Sex Differences in the Cancer Proteome
Proteins play a central role in cancer biology: they are the most common drug targets and biomarkers. Sex influences the proteome in many diseases, ranging from neurological to cardiovascular. In cancer, sex is associated with incidence, progression and therapeutic response, as well as characteristics of the tumour genome and transcriptome. The extent to which sex differences impact the cancer proteome remains largely unknown. To fill this gap, we quantified sex differences across 1,590 proteomes from eight cancer types, identifying 901 genes with sex-differential proteins abundance in adenocarcinomas of the lung, and 20 genes across five other tumour types: squamous cell carcinoma of the lung, hepatocellular carcinomas, clear cell cancers of the kidney, adenocarcinomas of the pancreas and glioblastoma. A subset of these protein differences could be rationalized by sex-differential copy number aberrations. Pathway analysis showed that male-biased proteins in lung adenocarcinoma were enriched in MYC and E2F target pathways. These findings highlight the modest impact of sex on the cancer proteome, but the very limited power of existing proteomics cohorts for these analyses.
Open AccessSex Differences in the Cancer Proteome
Proteins play a central role in cancer biology: they are the most common drug targets and biomarkers. Sex influences the proteome in many diseases, ranging from neurological to cardiovascular. In cancer, sex is associated with incidence, progression and therapeutic response, as well as characteristics of the tumour genome and transcriptome. The extent to which sex differences impact the cancer proteome remains largely unknown. To fill this gap, we quantified sex differences across 1,590 proteomes from eight cancer types, identifying 901 genes with sex-differential proteins abundance in adenocarcinomas of the lung, and 20 genes across five other tumour types: squamous cell carcinoma of the lung, hepatocellular carcinomas, clear cell cancers of the kidney, adenocarcinomas of the pancreas and glioblastoma. A subset of these protein differences could be rationalized by sex-differential copy number aberrations. Pathway analysis showed that male-biased proteins in lung adenocarcinoma were enriched in MYC and E2F target pathways. These findings highlight the modest impact of sex on the cancer proteome, but the very limited power of existing proteomics cohorts for these analyses.
- Open Access
Sex Differences in Cancer Functional Genomics: Gene Dependency and Drug Sensitivity
Patient sex influences a wide range of cancer phenotypes, including prevalence, response to therapy and survival endpoints. Molecular sex differences across the central dogma have been identified that may drive these phenotypic differences. Despite a growing catalog of specific genomic, transcriptomic and proteomic sex differences in a range of cancer types, their functional consequences remain unclear. To assess how patient sex impacts cancer cell function, we evaluated 1,209 cell lines subjected to CRISPR knockout, RNAi knockdown or drug exposures. Despite limited statistical power, we identified pan- and per-cancer sex differences in gene essentiality in six sex-linked and fourteen autosomal genes, and in drug sensitivity for two compounds. These data expand our understanding of the propagation of sex-differing molecular features to functional outcomes, and their role in influencing patient phenotypes. They highlight the importance of considering sex-specific effects in mechanistic and functional studies.
Open AccessSex Differences in Cancer Functional Genomics: Gene Dependency and Drug Sensitivity
Patient sex influences a wide range of cancer phenotypes, including prevalence, response to therapy and survival endpoints. Molecular sex differences across the central dogma have been identified that may drive these phenotypic differences. Despite a growing catalog of specific genomic, transcriptomic and proteomic sex differences in a range of cancer types, their functional consequences remain unclear. To assess how patient sex impacts cancer cell function, we evaluated 1,209 cell lines subjected to CRISPR knockout, RNAi knockdown or drug exposures. Despite limited statistical power, we identified pan- and per-cancer sex differences in gene essentiality in six sex-linked and fourteen autosomal genes, and in drug sensitivity for two compounds. These data expand our understanding of the propagation of sex-differing molecular features to functional outcomes, and their role in influencing patient phenotypes. They highlight the importance of considering sex-specific effects in mechanistic and functional studies.
- Open Access
- 1 May 2024
- Published by American Society of Clinical Oncology (ASCO) in JCO Clinical Cancer Informatics
- Vol. 8 (8) , e2300122
Use of Natural Language Processing to Infer Sites of Metastatic Disease From Radiology Reports at Scale
PURPOSE: To evaluate natural language processing (NLP) methods to infer metastatic sites from radiology reports. METHODS: A set of 4,522 computed tomography (CT) reports of 550 patients with 14 types of cancer was used to fine-tune four clinical large language models (LLMs) for multilabel classification of metastatic sites. We also developed an NLP information extraction (IE) system (on the basis of named entity recognition, assertion status detection, and relation extraction) for comparison. Model performances were measured by F1 scores on test and three external validation sets. The best model was used to facilitate analysis of metastatic frequencies in a cohort study of 6,555 patients with 53,838 CT reports. RESULTS: The RadBERT, BioBERT, GatorTron-base, and GatorTron-medium LLMs achieved F1 scores of 0.84, 0.87, 0.89, and 0.91, respectively, on the test set. The IE system performed best, achieving an F1 score of 0.93. F1 scores of the IE system by individual cancer type ranged from 0.89 to 0.96. The IE system attained F1 scores of 0.89, 0.83, and 0.81, respectively, on external validation sets including additional cancer types, positron emission tomography-CT ,and magnetic resonance imaging scans, respectively. In our cohort study, we found that for colorectal cancer, liver-only metastases were higher in de novo stage IV versus recurrent patients (29.7% v 12.2%; P < .001). Conversely, lung-only metastases were more frequent in recurrent versus de novo stage IV patients (17.2% v 7.3%; P < .001). CONCLUSION: We developed an IE system that accurately infers metastatic sites in multiple primary cancers from radiology reports. It has explainable methods and performs better than some clinical LLMs. The inferred metastatic phenotypes could enhance cancer research databases and clinical trial matching, and identify potential patients for oligometastatic interventions.
- 1 May 2024
- Published by American Society of Clinical Oncology (ASCO) in JCO Clinical Cancer Informatics
- Vol. 8 (8) , e2300122
Open AccessUse of Natural Language Processing to Infer Sites of Metastatic Disease From Radiology Reports at Scale
PURPOSE: To evaluate natural language processing (NLP) methods to infer metastatic sites from radiology reports. METHODS: A set of 4,522 computed tomography (CT) reports of 550 patients with 14 types of cancer was used to fine-tune four clinical large language models (LLMs) for multilabel classification of metastatic sites. We also developed an NLP information extraction (IE) system (on the basis of named entity recognition, assertion status detection, and relation extraction) for comparison. Model performances were measured by F1 scores on test and three external validation sets. The best model was used to facilitate analysis of metastatic frequencies in a cohort study of 6,555 patients with 53,838 CT reports. RESULTS: The RadBERT, BioBERT, GatorTron-base, and GatorTron-medium LLMs achieved F1 scores of 0.84, 0.87, 0.89, and 0.91, respectively, on the test set. The IE system performed best, achieving an F1 score of 0.93. F1 scores of the IE system by individual cancer type ranged from 0.89 to 0.96. The IE system attained F1 scores of 0.89, 0.83, and 0.81, respectively, on external validation sets including additional cancer types, positron emission tomography-CT ,and magnetic resonance imaging scans, respectively. In our cohort study, we found that for colorectal cancer, liver-only metastases were higher in de novo stage IV versus recurrent patients (29.7% v 12.2%; P < .001). Conversely, lung-only metastases were more frequent in recurrent versus de novo stage IV patients (17.2% v 7.3%; P < .001). CONCLUSION: We developed an IE system that accurately infers metastatic sites in multiple primary cancers from radiology reports. It has explainable methods and performs better than some clinical LLMs. The inferred metastatic phenotypes could enhance cancer research databases and clinical trial matching, and identify potential patients for oligometastatic interventions.
- Open Access
- 8 December 2022
- Published by Springer Nature in Nature Communications
- Vol. 13 (1) , 1
Author Correction: Divergent mutational processes distinguish hypoxic and normoxic tumours
- 8 December 2022
- Published by Springer Nature in Nature Communications
- Vol. 13 (1) , 1
Open AccessAuthor Correction: Divergent mutational processes distinguish hypoxic and normoxic tumours
- Open Access
- 28 August 2020
- Published by Springer Nature in Nature Communications
- Vol. 11 (1) , 1-24
Sex differences in oncogenic mutational processes
Sex differences have been observed in multiple facets of cancer epidemiology, treatment and biology, and in most cancers outside the sex organs. Efforts to link these clinical differences to specific molecular features have focused on somatic mutations within the coding regions of the genome. Here we report a pan-cancer analysis of sex differences in whole genomes of 1983 tumours of 28 subtypes as part of the ICGC/TCGA Pan-Cancer Analysis of Whole Genomes (PCAWG) Consortium. We both confirm the results of exome studies, and also uncover previously undescribed sex differences. These include sex-biases in coding and non-coding cancer drivers, mutation prevalence and strikingly, in mutational signatures related to underlying mutational processes. These results underline the pervasiveness of molecular sex differences and strengthen the call for increased consideration of sex in molecular cancer research.
- 28 August 2020
- Published by Springer Nature in Nature Communications
- Vol. 11 (1) , 1-24
Open AccessSex differences in oncogenic mutational processes
Sex differences have been observed in multiple facets of cancer epidemiology, treatment and biology, and in most cancers outside the sex organs. Efforts to link these clinical differences to specific molecular features have focused on somatic mutations within the coding regions of the genome. Here we report a pan-cancer analysis of sex differences in whole genomes of 1983 tumours of 28 subtypes as part of the ICGC/TCGA Pan-Cancer Analysis of Whole Genomes (PCAWG) Consortium. We both confirm the results of exome studies, and also uncover previously undescribed sex differences. These include sex-biases in coding and non-coding cancer drivers, mutation prevalence and strikingly, in mutational signatures related to underlying mutational processes. These results underline the pervasiveness of molecular sex differences and strengthen the call for increased consideration of sex in molecular cancer research.
- Open Access
Ancestry Influences on the Molecular Presentation of Tumours
Epidemiological studies have identified innumerable ways in which cancer presentation and behaviour is associated with patient ancestry. The molecular bases for these relationships remain largely unknown. We analyzed ancestry associations in the somatic mutational landscape of 12,774 tumours across 33 tumour-types, including 2,562 with whole-genome sequencing. Ancestry influences both the number of mutations in a tumour and the evolutionary timing of when they occur. Specific mutational signatures are associated with ancestry, reflecting potential differences in exogenous and endogenous oncogenic processes. A subset of known cancer driver genes was mutated in ancestry-associated patterns, with transcriptomic consequences. Cancer genome sequencing data is not well-balanced in epidemiologic factors; these data suggest ancestry strongly shapes the somatic mutational landscape of cancer, with potential functional implications.
Open AccessAncestry Influences on the Molecular Presentation of Tumours
Epidemiological studies have identified innumerable ways in which cancer presentation and behaviour is associated with patient ancestry. The molecular bases for these relationships remain largely unknown. We analyzed ancestry associations in the somatic mutational landscape of 12,774 tumours across 33 tumour-types, including 2,562 with whole-genome sequencing. Ancestry influences both the number of mutations in a tumour and the evolutionary timing of when they occur. Specific mutational signatures are associated with ancestry, reflecting potential differences in exogenous and endogenous oncogenic processes. A subset of known cancer driver genes was mutated in ancestry-associated patterns, with transcriptomic consequences. Cancer genome sequencing data is not well-balanced in epidemiologic factors; these data suggest ancestry strongly shapes the somatic mutational landscape of cancer, with potential functional implications.
- Open Access
Age Influences on the Molecular Presentation of Tumours
Cancer is often called a disease of aging. There are numerous ways in which cancer epidemiology and behaviour change with the age of the patient. The molecular bases for these relationships remain largely underexplored. To characterize them, we analyzed age-associations in the nuclear and mitochondrial somatic mutational landscape of 20,033 tumours across 35 tumour-types. Age influences both the number of mutations in a tumour and their evolutionary timing. Specific mutational signatures are associated with age, reflecting differences in exogenous and endogenous oncogenic processes such as a greater influence of tobacco use in the tumours of younger patients, but more activity of DNA damage repair signatures in those of older patients. We find that known cancer driver genes are mutated in age-associated frequencies, and these alter the transcriptome and predict for clinical outcomes. These effects are most striking in brain cancers where alterations like SUFU loss and ATRX mutation are age-dependent prognostic biomarkers. Using three cancer datasets, we show that age shapes the somatic mutational landscape of cancer, with clinical implications.
Open AccessAge Influences on the Molecular Presentation of Tumours
Cancer is often called a disease of aging. There are numerous ways in which cancer epidemiology and behaviour change with the age of the patient. The molecular bases for these relationships remain largely underexplored. To characterize them, we analyzed age-associations in the nuclear and mitochondrial somatic mutational landscape of 20,033 tumours across 35 tumour-types. Age influences both the number of mutations in a tumour and their evolutionary timing. Specific mutational signatures are associated with age, reflecting differences in exogenous and endogenous oncogenic processes such as a greater influence of tobacco use in the tumours of younger patients, but more activity of DNA damage repair signatures in those of older patients. We find that known cancer driver genes are mutated in age-associated frequencies, and these alter the transcriptome and predict for clinical outcomes. These effects are most striking in brain cancers where alterations like SUFU loss and ATRX mutation are age-dependent prognostic biomarkers. Using three cancer datasets, we show that age shapes the somatic mutational landscape of cancer, with clinical implications.
- Open Access
- 5 February 2020
- Published by Springer Nature in Nature Communications
- Vol. 11 (1) , 1-10
Divergent mutational processes distinguish hypoxic and normoxic tumours
Many primary tumours have low levels of molecular oxygen (hypoxia), and hypoxic tumours respond poorly to therapy. Pan-cancer molecular hallmarks of tumour hypoxia remain poorly understood, with limited comprehension of its associations with specific mutational processes, non-coding driver genes and evolutionary features. Here, as part of the ICGC/TCGA Pan-Cancer Analysis of Whole Genomes (PCAWG) Consortium, which aggregated whole genome sequencing data from 2658 cancers across 38 tumour types, we quantify hypoxia in 1188 tumours spanning 27 cancer types. Elevated hypoxia associates with increased mutational load across cancer types, irrespective of underlying mutational class. The proportion of mutations attributed to several mutational signatures of unknown aetiology directly associates with the level of hypoxia, suggesting underlying mutational processes for these signatures. At the gene level, driver mutations inTP53,MYCandPTENare enriched in hypoxic tumours, and mutations inPTENinteract with hypoxia to direct tumour evolutionary trajectories. Overall, hypoxia plays a critical role in shaping the genomic and evolutionary landscapes of cancer.
- 5 February 2020
- Published by Springer Nature in Nature Communications
- Vol. 11 (1) , 1-10
Open AccessDivergent mutational processes distinguish hypoxic and normoxic tumours
Many primary tumours have low levels of molecular oxygen (hypoxia), and hypoxic tumours respond poorly to therapy. Pan-cancer molecular hallmarks of tumour hypoxia remain poorly understood, with limited comprehension of its associations with specific mutational processes, non-coding driver genes and evolutionary features. Here, as part of the ICGC/TCGA Pan-Cancer Analysis of Whole Genomes (PCAWG) Consortium, which aggregated whole genome sequencing data from 2658 cancers across 38 tumour types, we quantify hypoxia in 1188 tumours spanning 27 cancer types. Elevated hypoxia associates with increased mutational load across cancer types, irrespective of underlying mutational class. The proportion of mutations attributed to several mutational signatures of unknown aetiology directly associates with the level of hypoxia, suggesting underlying mutational processes for these signatures. At the gene level, driver mutations inTP53,MYCandPTENare enriched in hypoxic tumours, and mutations inPTENinteract with hypoxia to direct tumour evolutionary trajectories. Overall, hypoxia plays a critical role in shaping the genomic and evolutionary landscapes of cancer.
- Open Access
- 28 January 2019
- Published by Cold Spring Harbor Laboratory in bioRxiv
- p. 531996
Divergent mutational processes distinguish hypoxic and normoxic tumours
Many primary tumours have low levels of molecular oxygen (hypoxia). Hypoxic tumours are more likely to metastasize to distant sites and respond poorly to multiple therapies. Surprisingly, then, the pan-cancer molecular hallmarks of tumour hypoxia remain poorly understood, with limited understanding of its associations with specific mutational processes, non-coding driver genes and evolutionary features. To fill this gap, we quantified hypoxia in 1,188 tumours spanning 27 cancer types. We show that elevated hypoxia is associated with increased mutational load across cancers, irrespective of the underlying mutational class. The proportion of mutations attributed to several mutational signatures of unknown aetiology are directly associated with the level of hypoxia, suggesting underlying mutational processes for these signatures. At the gene level, driver mutations in TP53, MYC and PTEN are enriched in tumours with high hypoxia, and mutations in PTEN interact with hypoxia to direct the evolutionary trajectory of tumours. Overall, this work demonstrates that hypoxia plays a critical role in shaping the genomic landscape of cancer.
- 28 January 2019
- Published by Cold Spring Harbor Laboratory in bioRxiv
- p. 531996
Open AccessDivergent mutational processes distinguish hypoxic and normoxic tumours
Many primary tumours have low levels of molecular oxygen (hypoxia). Hypoxic tumours are more likely to metastasize to distant sites and respond poorly to multiple therapies. Surprisingly, then, the pan-cancer molecular hallmarks of tumour hypoxia remain poorly understood, with limited understanding of its associations with specific mutational processes, non-coding driver genes and evolutionary features. To fill this gap, we quantified hypoxia in 1,188 tumours spanning 27 cancer types. We show that elevated hypoxia is associated with increased mutational load across cancers, irrespective of the underlying mutational class. The proportion of mutations attributed to several mutational signatures of unknown aetiology are directly associated with the level of hypoxia, suggesting underlying mutational processes for these signatures. At the gene level, driver mutations in TP53, MYC and PTEN are enriched in tumours with high hypoxia, and mutations in PTEN interact with hypoxia to direct the evolutionary trajectory of tumours. Overall, this work demonstrates that hypoxia plays a critical role in shaping the genomic landscape of cancer.