Daniel K. Putnam
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Daniel K. Putnam has a total of 172 co-authors. Here are the top co-authors by number of shared publications:
- Xiang Chen
Department of Computational Biology, St. Jude Children’s Research Hospital, Memphis, USA
11 shared publications - Jinghui Zhang
St Jude Children's Research Hospital, Memphis, TN
7 shared publications - Xiaotu Ma
St Jude Children's Research Hospital, Memphis, Tennessee, United States
7 shared publications - Benjamin J. Raphael
Department of Computer Science, Princeton University, Princeton, USA
6 shared publications John Easton
6 shared publications- Michael N. Edmonson
St. Jude Children’s Research Hospital 2 Department of Computational Biology Memphis, Tennessee
6 shared publications
Latest Publications
- Open Access
- 2 November 2026
- Published by American Society of Hematology in Blood
Identifying novel genomic alterations in relapsed childhood B-ALL: a report from the Children's Oncology Group
Relapsed B-acute lymphoblastic leukemia (B-ALL) generally occurs due to the emergence of subclones resistant to therapy, leading to poor outcome. Here, we describe novel genomic alterations responsible for therapeutic resistance at a subtype level in a clinical trial (AALL1331) which randomized patients to blinatumomab and chemotherapy versus chemotherapy alone. We performed whole-genome and transcriptome sequencing on 288 patients (183 diagnosis/first relapse/germline trios and 105 additional patients with first-relapse samples) and analyzed 41 second-relapse samples. For ETV6::RUNX1, we showed that cases with gain of the fusion allele or with fusion transcript involving exon 5 of ETV6 and exon 4 of RUNX1 (E5-E4) had a higher risk of first relapse than cases with E5-E3. For the hyperdiploid subtype, we demonstrated that low modal chromosomal numbers (51-52) were more prone to first relapse than those with ≥53 chromosomes, with an additive effect of CREBBP mutations. We showed enrichment of IKZF1 alterations in the Philadelphia-like subtype and that NSD2 E1099K was commonly associated with relapse in TCF3::PBX1. Together, these new features accounted for 30% (n=55) of first relapses. Intriguingly, ~3% of relapses are related to secondary B-ALL or germline predisposition. We also compared genomic alterations at diagnosis and first relapse in relation to second relapse status in randomized patients, with TCF3::PBX1/NT5C2 linked to worse outcomes regardless of randomized arm. Together, our findings provide novel insights into the genomics of relapsed childhood B-ALL as well as in the context of blinatumomab therapy. This trial was registered at www.clinicaltrials.gov as # NCT0201853.
- 2 November 2026
- Published by American Society of Hematology in Blood
Open AccessIdentifying novel genomic alterations in relapsed childhood B-ALL: a report from the Children's Oncology Group
Relapsed B-acute lymphoblastic leukemia (B-ALL) generally occurs due to the emergence of subclones resistant to therapy, leading to poor outcome. Here, we describe novel genomic alterations responsible for therapeutic resistance at a subtype level in a clinical trial (AALL1331) which randomized patients to blinatumomab and chemotherapy versus chemotherapy alone. We performed whole-genome and transcriptome sequencing on 288 patients (183 diagnosis/first relapse/germline trios and 105 additional patients with first-relapse samples) and analyzed 41 second-relapse samples. For ETV6::RUNX1, we showed that cases with gain of the fusion allele or with fusion transcript involving exon 5 of ETV6 and exon 4 of RUNX1 (E5-E4) had a higher risk of first relapse than cases with E5-E3. For the hyperdiploid subtype, we demonstrated that low modal chromosomal numbers (51-52) were more prone to first relapse than those with ≥53 chromosomes, with an additive effect of CREBBP mutations. We showed enrichment of IKZF1 alterations in the Philadelphia-like subtype and that NSD2 E1099K was commonly associated with relapse in TCF3::PBX1. Together, these new features accounted for 30% (n=55) of first relapses. Intriguingly, ~3% of relapses are related to secondary B-ALL or germline predisposition. We also compared genomic alterations at diagnosis and first relapse in relation to second relapse status in randomized patients, with TCF3::PBX1/NT5C2 linked to worse outcomes regardless of randomized arm. Together, our findings provide novel insights into the genomics of relapsed childhood B-ALL as well as in the context of blinatumomab therapy. This trial was registered at www.clinicaltrials.gov as # NCT0201853.
- Open Access
- 1 May 2026
- Published by American Association for the Advancement of Science (AAAS) in Science Advances
- Vol. 12 (18) , eaeb3503
IRS4 is a PI3K-activating cancer dependency up-regulated through DNA rearrangements or epigenetic mechanisms in multiple solid tumors
Cancer therapeutics frequently fail in clinical trials because of poor therapeutic index (efficacy-to-toxicity ratio). We systematically identified targets likely to have a good therapeutic index, revealing insulin receptor substrate 4 (IRS4) as a dependency in IRS4 -expressing cancers. Pan-cancer analysis of pediatric-enriched cancers revealed IRS4 expression consistent with dependency in 68% of choroid plexus, 37% of malignant rhabdoid, 31% of NUT midline, and 5% of osteosarcomas, while in adult cancers, it was expressed in 8% of uterine leiomyosarcomas and 1 to 2% of lung squamous, stomach, and breast carcinomas. IRS4 expression in adult tumors was associated with enhancer hijacking rearrangements, including recurrent GATA3-IRS4 and ANKRD30A-IRS4 in breast cancer, while rhabdoid and NUT midline cancers expressed IRS4 epigenetically. IRS4 fueled cancer dependency through PI3K-Akt activation, and domain analysis revealed the PH and PTB domains, which have a predicted drug pocket, to be dispensable, suggesting degradation-based modalities. These data reveal IRS4 as a target in IRS4-expressing cancers and suggest inhibitory approaches.
- 1 May 2026
- Published by American Association for the Advancement of Science (AAAS) in Science Advances
- Vol. 12 (18) , eaeb3503
Open AccessIRS4 is a PI3K-activating cancer dependency up-regulated through DNA rearrangements or epigenetic mechanisms in multiple solid tumors
Cancer therapeutics frequently fail in clinical trials because of poor therapeutic index (efficacy-to-toxicity ratio). We systematically identified targets likely to have a good therapeutic index, revealing insulin receptor substrate 4 (IRS4) as a dependency in IRS4 -expressing cancers. Pan-cancer analysis of pediatric-enriched cancers revealed IRS4 expression consistent with dependency in 68% of choroid plexus, 37% of malignant rhabdoid, 31% of NUT midline, and 5% of osteosarcomas, while in adult cancers, it was expressed in 8% of uterine leiomyosarcomas and 1 to 2% of lung squamous, stomach, and breast carcinomas. IRS4 expression in adult tumors was associated with enhancer hijacking rearrangements, including recurrent GATA3-IRS4 and ANKRD30A-IRS4 in breast cancer, while rhabdoid and NUT midline cancers expressed IRS4 epigenetically. IRS4 fueled cancer dependency through PI3K-Akt activation, and domain analysis revealed the PH and PTB domains, which have a predicted drug pocket, to be dispensable, suggesting degradation-based modalities. These data reveal IRS4 as a target in IRS4-expressing cancers and suggest inhibitory approaches.
- Open Access
- 29 November 2025
- Published by Oxford University Press (OUP) in Genomics, Proteomics & Bioinformatics
- Vol. 23 (6)
CanID: A Robust and Accurate RNA-seq Expression-based Diagnostic Classification Scheme for Pediatric Malignancies
Cancer subtype classification is critical for precision therapy, and there is a growing trend to augment histopathology testing with omics-based machine learning classifiers. However, analytical challenges remain in pediatric cancer regarding the scope and precision of current classifiers, as well as the evolving subtype standardization. To address these challenges, we constructed Cancer Identification (CanID), a stacked ensemble machine learning classification scheme, using transcriptomic features derived from gene-level RNA sequencing count data as the sole input. CanID was developed primarily from 3203 pediatric cancer samples across 13 solid tumor subtypes and 38 hematologic malignancy subtypes, with subtype labels curated without the use of RNA-seq data. The accuracies of independent testing in three independent or external datasets for solid tumors and hematologic malignancies were 99% and 92%–93%, respectively. Notably, CanID was able to classify subtypes challenging for clinical histology evaluation and was robust to both biological and technical challenges, including differences in data collection protocols, class imbalance, potential mislabeled training samples, and classes unobserved during training. The high accuracy, robustness, and biological interpretability of this transcriptome-based classification scheme represent a valuable approach to advance tumor diagnosis and clinically meaningful stratification of tumor types. CanID can be accessed on GitHub at https://github.com/chenlab-sj/CanID.
- 29 November 2025
- Published by Oxford University Press (OUP) in Genomics, Proteomics & Bioinformatics
- Vol. 23 (6)
Open AccessCanID: A Robust and Accurate RNA-seq Expression-based Diagnostic Classification Scheme for Pediatric Malignancies
Cancer subtype classification is critical for precision therapy, and there is a growing trend to augment histopathology testing with omics-based machine learning classifiers. However, analytical challenges remain in pediatric cancer regarding the scope and precision of current classifiers, as well as the evolving subtype standardization. To address these challenges, we constructed Cancer Identification (CanID), a stacked ensemble machine learning classification scheme, using transcriptomic features derived from gene-level RNA sequencing count data as the sole input. CanID was developed primarily from 3203 pediatric cancer samples across 13 solid tumor subtypes and 38 hematologic malignancy subtypes, with subtype labels curated without the use of RNA-seq data. The accuracies of independent testing in three independent or external datasets for solid tumors and hematologic malignancies were 99% and 92%–93%, respectively. Notably, CanID was able to classify subtypes challenging for clinical histology evaluation and was robust to both biological and technical challenges, including differences in data collection protocols, class imbalance, potential mislabeled training samples, and classes unobserved during training. The high accuracy, robustness, and biological interpretability of this transcriptome-based classification scheme represent a valuable approach to advance tumor diagnosis and clinically meaningful stratification of tumor types. CanID can be accessed on GitHub at https://github.com/chenlab-sj/CanID.
- Open Access
CanID: a robust and accurate RNAseq Expression-based diagnostic classification scheme for pediatric malignancies
Cancer subtype classification is critical for precision therapy and there is a growing trend of augmenting histopathology testing procedures with omics-based machine learning classifiers. However, analytical challenges remain for pediatric cancer on the scope and precision of the current classifiers as well as the evolving subtype standardization. To address these challenges, we built Cancer Identification or CanID, a stacked ensemble machine learning classification scheme, using the transcriptomic features derived from gene-level RNA sequencing count data as the sole input. CanID was developed primarily from 3203 pediatric cancer samples of 13 solid tumor subtypes and 38 hematologic malignancy subtypes with subtype labels curated without the use of RNA-seq data. The accuracies of independent testing in three independent or external data sets for Solid Tumor and Hematologic Malignancy are 99% and 92–93%, respectively. Notably, CanID was able to classify subtypes challenging for clinical histology evaluation and was robust to both biological and technical challenges, including differences in data collection protocols, class imbalance, potential mislabeled training samples and classes unobserved in training. The high accuracy, robustness, biological interpretability of this transcriptome-based classification scheme represents a valuable approach to advance tumor diagnosis and clinically meaningful stratification of tumor types. CanID can be accessed on GitHub at https://github.com/chenlab-sj/CanID .
Open AccessCanID: a robust and accurate RNAseq Expression-based diagnostic classification scheme for pediatric malignancies
Cancer subtype classification is critical for precision therapy and there is a growing trend of augmenting histopathology testing procedures with omics-based machine learning classifiers. However, analytical challenges remain for pediatric cancer on the scope and precision of the current classifiers as well as the evolving subtype standardization. To address these challenges, we built Cancer Identification or CanID, a stacked ensemble machine learning classification scheme, using the transcriptomic features derived from gene-level RNA sequencing count data as the sole input. CanID was developed primarily from 3203 pediatric cancer samples of 13 solid tumor subtypes and 38 hematologic malignancy subtypes with subtype labels curated without the use of RNA-seq data. The accuracies of independent testing in three independent or external data sets for Solid Tumor and Hematologic Malignancy are 99% and 92–93%, respectively. Notably, CanID was able to classify subtypes challenging for clinical histology evaluation and was robust to both biological and technical challenges, including differences in data collection protocols, class imbalance, potential mislabeled training samples and classes unobserved in training. The high accuracy, robustness, biological interpretability of this transcriptome-based classification scheme represents a valuable approach to advance tumor diagnosis and clinically meaningful stratification of tumor types. CanID can be accessed on GitHub at https://github.com/chenlab-sj/CanID .
- Open Access
- 3 April 2023
- Published by American Association for Cancer Research (AACR)
Supplementary Figures S1-S11 from The Clonal Evolution of Metastatic Osteosarcoma as Shaped by Cisplatin Treatment
Supplementary Figures S1-S11 show sample histology, quality control data, and additional genomic analysis. S1. Histology of osteosarcoma SJOS001101. S2. Whole-genome sequencing coverage metrics. S3. Deep capture validation coverage metrics. S4. Example genes experiencing copy number alterations. S5. B-allele frequencies confirm homozygous deletions of 3q13.31 and CDKN2A. S6. Identification of cisplatin mutation signature in osteosarcoma. S7. Cisplatin does not induce kataegis. S8. Density plot analysis of SNV clusters reveals unique clones at most sites. S9. SJOS001101 pairwise osteosarcoma sample SNV comparisons reveal cross-seeding. S10. Clonal evolution of SJOS001101 as clarified by MACHINA analysis. S11. Pairwise osteosarcoma sample SNV comparisons in three additional patients.
- 3 April 2023
- Published by American Association for Cancer Research (AACR)
Open AccessSupplementary Figures S1-S11 from The Clonal Evolution of Metastatic Osteosarcoma as Shaped by Cisplatin Treatment
Supplementary Figures S1-S11 show sample histology, quality control data, and additional genomic analysis. S1. Histology of osteosarcoma SJOS001101. S2. Whole-genome sequencing coverage metrics. S3. Deep capture validation coverage metrics. S4. Example genes experiencing copy number alterations. S5. B-allele frequencies confirm homozygous deletions of 3q13.31 and CDKN2A. S6. Identification of cisplatin mutation signature in osteosarcoma. S7. Cisplatin does not induce kataegis. S8. Density plot analysis of SNV clusters reveals unique clones at most sites. S9. SJOS001101 pairwise osteosarcoma sample SNV comparisons reveal cross-seeding. S10. Clonal evolution of SJOS001101 as clarified by MACHINA analysis. S11. Pairwise osteosarcoma sample SNV comparisons in three additional patients.
- Open Access
- 3 April 2023
- Published by American Association for Cancer Research (AACR)
Supplementary Tables S1-S5 from The Clonal Evolution of Metastatic Osteosarcoma as Shaped by Cisplatin Treatment
Osteosarcoma clinical history and capture validation sequencing data. Supplementary Table S1. Sample and treatment information for relapsed osteosarcoma patients. Supplementary Table S2. SJOS0011101 SNVs detected by capture validation and tumor purity. Supplementary Table S3. SJOS0011105 SNVs detected by capture validation and tumor purity. Supplementary Table S4. SJOS0011107 SNVs detected by capture validation and tumor purity. Supplementary Table S5. SJOS010 SNVs detected by capture validation and tumor purity.
- 3 April 2023
- Published by American Association for Cancer Research (AACR)
Open AccessSupplementary Tables S1-S5 from The Clonal Evolution of Metastatic Osteosarcoma as Shaped by Cisplatin Treatment
Osteosarcoma clinical history and capture validation sequencing data. Supplementary Table S1. Sample and treatment information for relapsed osteosarcoma patients. Supplementary Table S2. SJOS0011101 SNVs detected by capture validation and tumor purity. Supplementary Table S3. SJOS0011105 SNVs detected by capture validation and tumor purity. Supplementary Table S4. SJOS0011107 SNVs detected by capture validation and tumor purity. Supplementary Table S5. SJOS010 SNVs detected by capture validation and tumor purity.
- Open Access
- 3 April 2023
- Published by American Association for Cancer Research (AACR)
Supplementary Figures S1-S11 from The Clonal Evolution of Metastatic Osteosarcoma as Shaped by Cisplatin Treatment
Supplementary Figures S1-S11 show sample histology, quality control data, and additional genomic analysis. S1. Histology of osteosarcoma SJOS001101. S2. Whole-genome sequencing coverage metrics. S3. Deep capture validation coverage metrics. S4. Example genes experiencing copy number alterations. S5. B-allele frequencies confirm homozygous deletions of 3q13.31 and CDKN2A. S6. Identification of cisplatin mutation signature in osteosarcoma. S7. Cisplatin does not induce kataegis. S8. Density plot analysis of SNV clusters reveals unique clones at most sites. S9. SJOS001101 pairwise osteosarcoma sample SNV comparisons reveal cross-seeding. S10. Clonal evolution of SJOS001101 as clarified by MACHINA analysis. S11. Pairwise osteosarcoma sample SNV comparisons in three additional patients.
- 3 April 2023
- Published by American Association for Cancer Research (AACR)
Open AccessSupplementary Figures S1-S11 from The Clonal Evolution of Metastatic Osteosarcoma as Shaped by Cisplatin Treatment
Supplementary Figures S1-S11 show sample histology, quality control data, and additional genomic analysis. S1. Histology of osteosarcoma SJOS001101. S2. Whole-genome sequencing coverage metrics. S3. Deep capture validation coverage metrics. S4. Example genes experiencing copy number alterations. S5. B-allele frequencies confirm homozygous deletions of 3q13.31 and CDKN2A. S6. Identification of cisplatin mutation signature in osteosarcoma. S7. Cisplatin does not induce kataegis. S8. Density plot analysis of SNV clusters reveals unique clones at most sites. S9. SJOS001101 pairwise osteosarcoma sample SNV comparisons reveal cross-seeding. S10. Clonal evolution of SJOS001101 as clarified by MACHINA analysis. S11. Pairwise osteosarcoma sample SNV comparisons in three additional patients.
- Open Access
- 3 April 2023
- Published by American Association for Cancer Research (AACR)
Data from The Clonal Evolution of Metastatic Osteosarcoma as Shaped by Cisplatin Treatment
To investigate the genomic evolution of metastatic pediatric osteosarcoma, we performed whole-genome and targeted deep sequencing on 14 osteosarcoma metastases and two primary tumors from four patients (two to eight samples per patient). All four patients harbored ancestral (truncal) somatic variants resulting in TP53 inactivation and cell-cycle aberrations, followed by divergence into relapse-specific lineages exhibiting a cisplatin-induced mutation signature. In three of the four patients, the cisplatin signature accounted for >40% of mutations detected in the metastatic samples. Mutations potentially acquired during cisplatin treatment included NF1 missense mutations of uncertain significance in two patients and a KIT G565R activating mutation in one patient. Three of four patients demonstrated widespread ploidy differences between samples from the sample patient. Single-cell seeding of metastasis was detected in most metastatic samples. Cross-seeding between metastatic sites was observed in one patient, whereas in another patient a minor clone from the primary tumor seeded both metastases analyzed. These results reveal extensive clonal heterogeneity in metastatic osteosarcoma, much of which is likely cisplatin-induced.Implications:The extent and consequences of chemotherapy-induced damage in pediatric cancers is unknown. We found that cisplatin treatment can potentially double the mutational burden in osteosarcoma, which has implications for optimizing therapy for recurrent, chemotherapy-resistant disease.
- 3 April 2023
- Published by American Association for Cancer Research (AACR)
Open AccessData from The Clonal Evolution of Metastatic Osteosarcoma as Shaped by Cisplatin Treatment
To investigate the genomic evolution of metastatic pediatric osteosarcoma, we performed whole-genome and targeted deep sequencing on 14 osteosarcoma metastases and two primary tumors from four patients (two to eight samples per patient). All four patients harbored ancestral (truncal) somatic variants resulting in TP53 inactivation and cell-cycle aberrations, followed by divergence into relapse-specific lineages exhibiting a cisplatin-induced mutation signature. In three of the four patients, the cisplatin signature accounted for >40% of mutations detected in the metastatic samples. Mutations potentially acquired during cisplatin treatment included NF1 missense mutations of uncertain significance in two patients and a KIT G565R activating mutation in one patient. Three of four patients demonstrated widespread ploidy differences between samples from the sample patient. Single-cell seeding of metastasis was detected in most metastatic samples. Cross-seeding between metastatic sites was observed in one patient, whereas in another patient a minor clone from the primary tumor seeded both metastases analyzed. These results reveal extensive clonal heterogeneity in metastatic osteosarcoma, much of which is likely cisplatin-induced.Implications:The extent and consequences of chemotherapy-induced damage in pediatric cancers is unknown. We found that cisplatin treatment can potentially double the mutational burden in osteosarcoma, which has implications for optimizing therapy for recurrent, chemotherapy-resistant disease.
- Open Access
- 3 April 2023
- Published by American Association for Cancer Research (AACR)
Supplementary Tables S1-S5 from The Clonal Evolution of Metastatic Osteosarcoma as Shaped by Cisplatin Treatment
Osteosarcoma clinical history and capture validation sequencing data. Supplementary Table S1. Sample and treatment information for relapsed osteosarcoma patients. Supplementary Table S2. SJOS0011101 SNVs detected by capture validation and tumor purity. Supplementary Table S3. SJOS0011105 SNVs detected by capture validation and tumor purity. Supplementary Table S4. SJOS0011107 SNVs detected by capture validation and tumor purity. Supplementary Table S5. SJOS010 SNVs detected by capture validation and tumor purity.
- 3 April 2023
- Published by American Association for Cancer Research (AACR)
Open AccessSupplementary Tables S1-S5 from The Clonal Evolution of Metastatic Osteosarcoma as Shaped by Cisplatin Treatment
Osteosarcoma clinical history and capture validation sequencing data. Supplementary Table S1. Sample and treatment information for relapsed osteosarcoma patients. Supplementary Table S2. SJOS0011101 SNVs detected by capture validation and tumor purity. Supplementary Table S3. SJOS0011105 SNVs detected by capture validation and tumor purity. Supplementary Table S4. SJOS0011107 SNVs detected by capture validation and tumor purity. Supplementary Table S5. SJOS010 SNVs detected by capture validation and tumor purity.
- Open Access
- 15 June 2022
- Published by American Association for Cancer Research (AACR) in Cancer Research
- Vol. 82 (12_Supplem) , 1936
Abstract 1936: MethylationToRegulation: A deep-learning approach to infer chromatin properties from DNA methylomes
Gene regulation is critical for cell identity, and its dysregulation is a defining characteristic of common diseases including cancers. Although promoter activity is a strong predictor of gene expression, activities at other regulatory elements, including enhancers and super-enhancers (SE), are major contributors to gene regulation. For example, distinct transcription factor-regulating SEs define neuroblastoma (NB) subtypes with distinct clinical outcomes. However, technical limitations largely prevent ChIP-seq based enhancer/SE activity profiling in primary patient samples. We previously developed MethylationToActivity (M2A) and demonstrated that high-order DNA methylation (DNAm) features are strong predictors of promoter activity measured by ChIP-seq. This tool is important because genomewide DNAm assays are widely used in clinic to classify tumors and stratify patients on clinical trials. However, it is limited by its reliance on known gene position annotations. Here we present MethylationToRegulation (M2R), a method that utilizes a convolutional neural network (CNN)-based deep learning framework to infer intergenic enhancer/SE activities from DNA methylomes. We obtained paired WGBS and H3K27ac ChIP-seq data for 16 pediatric NB samples profiled in the Pediatric Cancer Genome Project. M2R was trained on 6 samples and tested on the remaining 10 samples. It achieved an average relative prediction accuracy of 84% on test samples when compared to the H3K27ac ChIP-seq replicate consistency from the ENCODE project. We adapted the Rank Ordering of Super-Enhancers algorithm to interpret H3K27ac signals inferred from DNA methylomes to identify SEs. M2R faithfully captured subtype-defining SEs with high specificity, including those associated with master NB transcription factors. Our results demonstrate that M2R accurately quantifies enhancer/SE activities and infers critical epigenetic marks from DNA methylomes. Application of M2R will enable the profiling of epigenetic dysregulation in patient tumor samples, seeking to improve clinical outcomes. Citation Format: Daniel K. Putnam, Brian J. Abraham, Xiang Chen. MethylationToRegulation: A deep-learning approach to infer chromatin properties from DNA methylomes [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr 1936.
- 15 June 2022
- Published by American Association for Cancer Research (AACR) in Cancer Research
- Vol. 82 (12_Supplem) , 1936
Open AccessAbstract 1936: MethylationToRegulation: A deep-learning approach to infer chromatin properties from DNA methylomes
Gene regulation is critical for cell identity, and its dysregulation is a defining characteristic of common diseases including cancers. Although promoter activity is a strong predictor of gene expression, activities at other regulatory elements, including enhancers and super-enhancers (SE), are major contributors to gene regulation. For example, distinct transcription factor-regulating SEs define neuroblastoma (NB) subtypes with distinct clinical outcomes. However, technical limitations largely prevent ChIP-seq based enhancer/SE activity profiling in primary patient samples. We previously developed MethylationToActivity (M2A) and demonstrated that high-order DNA methylation (DNAm) features are strong predictors of promoter activity measured by ChIP-seq. This tool is important because genomewide DNAm assays are widely used in clinic to classify tumors and stratify patients on clinical trials. However, it is limited by its reliance on known gene position annotations. Here we present MethylationToRegulation (M2R), a method that utilizes a convolutional neural network (CNN)-based deep learning framework to infer intergenic enhancer/SE activities from DNA methylomes. We obtained paired WGBS and H3K27ac ChIP-seq data for 16 pediatric NB samples profiled in the Pediatric Cancer Genome Project. M2R was trained on 6 samples and tested on the remaining 10 samples. It achieved an average relative prediction accuracy of 84% on test samples when compared to the H3K27ac ChIP-seq replicate consistency from the ENCODE project. We adapted the Rank Ordering of Super-Enhancers algorithm to interpret H3K27ac signals inferred from DNA methylomes to identify SEs. M2R faithfully captured subtype-defining SEs with high specificity, including those associated with master NB transcription factors. Our results demonstrate that M2R accurately quantifies enhancer/SE activities and infers critical epigenetic marks from DNA methylomes. Application of M2R will enable the profiling of epigenetic dysregulation in patient tumor samples, seeking to improve clinical outcomes. Citation Format: Daniel K. Putnam, Brian J. Abraham, Xiang Chen. MethylationToRegulation: A deep-learning approach to infer chromatin properties from DNA methylomes [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr 1936.