Peter M. Clark
Research Subjects
Peter M. Clark has a total of 101 co-authors. Here are the top co-authors by number of shared publications:
- Eric Londin
Computational Medicine Center, Sidney Kimmel Medical College, Thomas Jefferson University, Philadelphia, PA
6 shared publications Paolo Fortina
Cancer Genomics Laboratory, Department of Cancer Biology, Kimmel Cancer Center, Sidney Kimmel Medical College at Thomas Jefferson University, Philadelphia, PA 19146, USA
4 shared publicationsDeborah Ferriola
3 shared publicationsDimitri S. Monos
3 shared publications- Dimitrios S. Monos
Department of Pathology and Laboratory Medicine, Children’s Hospital of Philadelphia, Philadelphia, PA, USA
3 shared publications - Isidore Rigoutsos
Computational Medicine Center, Thomas Jefferson University
3 shared publications
Latest Publications
- Open Access
EGGNet, a generalizable geometric deep learning framework for protein complex pose scoring
Computational prediction of molecule-protein interactions has been key for developing new molecules to interact with a target protein for therapeutics development. Past work includes two independent streams of approaches: (1) predicting protein-protein interactions (PPI) between naturally occurring proteins and (2) predicting the binding affinities between proteins and small molecule ligands (aka drug target interaction, or DTI). Studying the two problems in isolation has limited the ability of these computational models to generalize across the PPI and DTI tasks, both of which ultimately involve non-covalent interactions with a protein target. In this work, we developed an Equivariant Graph of Graphs neural Network (EGGNet), a geometric deep learning framework for molecule-protein binding predictions that can handle three types of molecules for interacting with a target protein: (1) small molecules, (2) synthetic peptides and (3) natural proteins. EGGNet leverages a graph of graphs (GoGs) representation constructed from the molecule structures at atomic-resolution and utilizes a multi-resolution equivariant graph neural network (GNN) to learn from such representations. In addition, EGGNet leverages the underlying biophysics and makes use of both atom- and residue-level interactions, which improve EGGNet’s ability to rank candidate poses from blind docking. EGGNet achieves competitive performance on both a public proteinsmall molecule binding affinity prediction task (80.2% top-1 success rate on CASF-2016) and an synthetic protein interface prediction task (88.4% AUPR). We envision that the proposed geometric deep learning framework can generalize to many other protein interaction prediction problems, such as binding site prediction and molecular docking, helping accelerate protein engineering and structure-based drug development.
Open AccessEGGNet, a generalizable geometric deep learning framework for protein complex pose scoring
Computational prediction of molecule-protein interactions has been key for developing new molecules to interact with a target protein for therapeutics development. Past work includes two independent streams of approaches: (1) predicting protein-protein interactions (PPI) between naturally occurring proteins and (2) predicting the binding affinities between proteins and small molecule ligands (aka drug target interaction, or DTI). Studying the two problems in isolation has limited the ability of these computational models to generalize across the PPI and DTI tasks, both of which ultimately involve non-covalent interactions with a protein target. In this work, we developed an Equivariant Graph of Graphs neural Network (EGGNet), a geometric deep learning framework for molecule-protein binding predictions that can handle three types of molecules for interacting with a target protein: (1) small molecules, (2) synthetic peptides and (3) natural proteins. EGGNet leverages a graph of graphs (GoGs) representation constructed from the molecule structures at atomic-resolution and utilizes a multi-resolution equivariant graph neural network (GNN) to learn from such representations. In addition, EGGNet leverages the underlying biophysics and makes use of both atom- and residue-level interactions, which improve EGGNet’s ability to rank candidate poses from blind docking. EGGNet achieves competitive performance on both a public proteinsmall molecule binding affinity prediction task (80.2% top-1 success rate on CASF-2016) and an synthetic protein interface prediction task (88.4% AUPR). We envision that the proposed geometric deep learning framework can generalize to many other protein interaction prediction problems, such as binding site prediction and molecular docking, helping accelerate protein engineering and structure-based drug development.
- Open Access
LM-GVP: A Generalizable Deep Learning Framework for Protein Property Prediction from Sequence and Structure
Proteins perform many essential functions in biological systems and can be successfully developed as bio-therapeutics. It is invaluable to be able to predict their properties based on a proposed sequence and structure. In this study, we developed a novel generalizable deep learning framework, LM-GVP, composed of a protein Language Model (LM) and Graph Neural Network (GNN) to leverage information from both 1D amino acid sequences and 3D structures of proteins. Our approach outperformed the state-of-the-art protein LMs on a variety of property prediction tasks including fluorescence, protease stability, and protein functions from Gene Ontology (GO). We also illustrated insights into how a GNN prediction head can guide the protein LM to better leverage structural information. We envision that our deep learning framework will be generalizable to many protein property prediction problems to greatly accelerate protein engineering and drug development.
Open AccessLM-GVP: A Generalizable Deep Learning Framework for Protein Property Prediction from Sequence and Structure
Proteins perform many essential functions in biological systems and can be successfully developed as bio-therapeutics. It is invaluable to be able to predict their properties based on a proposed sequence and structure. In this study, we developed a novel generalizable deep learning framework, LM-GVP, composed of a protein Language Model (LM) and Graph Neural Network (GNN) to leverage information from both 1D amino acid sequences and 3D structures of proteins. Our approach outperformed the state-of-the-art protein LMs on a variety of property prediction tasks including fluorescence, protease stability, and protein functions from Gene Ontology (GO). We also illustrated insights into how a GNN prediction head can guide the protein LM to better leverage structural information. We envision that our deep learning framework will be generalizable to many protein property prediction problems to greatly accelerate protein engineering and drug development.
- 15 February 2019
- Published by Elsevier
- p. 35-53
Computational methods for microRNA and PIWI-interacting RNA gene discovery and functional predictions
- 15 February 2019
- Published by Elsevier
- p. 35-53
Computational methods for microRNA and PIWI-interacting RNA gene discovery and functional predictions
- 1 September 2017
- Published by Elsevier in Human Immunology
- Vol. 78, 22
OR21 The novel functional role of HLA DRA and DRB5 encoded mirna transcripts
- 1 September 2017
- Published by Elsevier in Human Immunology
- Vol. 78, 22
OR21 The novel functional role of HLA DRA and DRB5 encoded mirna transcripts
- 1 September 2017
- Published by Elsevier in Human Immunology
- Vol. 78, 90
P050 Rapid, high resolution HLA genotyping using nanopore sequencing
- 1 September 2017
- Published by Elsevier in Human Immunology
- Vol. 78, 90
P050 Rapid, high resolution HLA genotyping using nanopore sequencing
- 1 September 2017
- Published by Elsevier in Human Immunology
- Vol. 78, 21
OR19 MIRNA mediated post transcriptional regulation of HLA transcripts
- 1 September 2017
- Published by Elsevier in Human Immunology
- Vol. 78, 21
OR19 MIRNA mediated post transcriptional regulation of HLA transcripts
- 1 December 2016
- Published by Oxford University Press (OUP) in Clinical Chemistry
- Vol. 62 (12) , 1630-1638
Generation of Full-Length Class I Human Leukocyte Antigen Gene Consensus Sequences for Novel Allele Characterization
BACKGROUND: Routine, high-resolution human leukocyte antigen (HLA) genotyping by next generation sequencing within clinical immunogenetics laboratories can now provide the full-length gene sequence characterization of fully phased HLA alleles. This powerful technique provides insights into HLA variation beyond the traditionally characterized antigen recognition domain, providing sequence annotation across the entire gene including untranslated and intronic regions and may be used to characterize novel alleles from massively parallel sequencing runs.METHODS: We evaluated the utility of the Omixon Holotype HLA assay to generate credible, fully phased full-length gene consensus sequences for 50 individuals at major histocompatibility complex, class I, A (HLA-A), HLA-B, and HLA-C loci (300 genotyped alleles in total) to identify and characterize novel class I HLA alleles using our downstream analytical pipeline.RESULTS: Our analysis revealed that 7.7% (23/300) of genotyped class I HLA alleles contain novel polymorphisms. Interestingly, all of the novel alleles identified by our analysis were found to harbor sequence variations within intronic regions of the respective locus. In total our analysis identified 17 unique novel class I HLA alleles from 23 of the 300 genotyped alleles and generated full-length gene sequence annotations for 9 previously incompletely annotated HLA class I allele sequences derived from 14 of the 300 genotyped alleles.CONCLUSIONS: The demonstrated utility of the Omixon Holotype HLA assay in combination with our downstream analytical framework to generate fully phased, full-length gene consensus sequences for the identification and characterization of novel HLA alleles, facilitates the study of HLA polymorphism beyond the antigen recognition domain in human health and disease.
- 1 December 2016
- Published by Oxford University Press (OUP) in Clinical Chemistry
- Vol. 62 (12) , 1630-1638
Generation of Full-Length Class I Human Leukocyte Antigen Gene Consensus Sequences for Novel Allele Characterization
BACKGROUND: Routine, high-resolution human leukocyte antigen (HLA) genotyping by next generation sequencing within clinical immunogenetics laboratories can now provide the full-length gene sequence characterization of fully phased HLA alleles. This powerful technique provides insights into HLA variation beyond the traditionally characterized antigen recognition domain, providing sequence annotation across the entire gene including untranslated and intronic regions and may be used to characterize novel alleles from massively parallel sequencing runs.METHODS: We evaluated the utility of the Omixon Holotype HLA assay to generate credible, fully phased full-length gene consensus sequences for 50 individuals at major histocompatibility complex, class I, A (HLA-A), HLA-B, and HLA-C loci (300 genotyped alleles in total) to identify and characterize novel class I HLA alleles using our downstream analytical pipeline.RESULTS: Our analysis revealed that 7.7% (23/300) of genotyped class I HLA alleles contain novel polymorphisms. Interestingly, all of the novel alleles identified by our analysis were found to harbor sequence variations within intronic regions of the respective locus. In total our analysis identified 17 unique novel class I HLA alleles from 23 of the 300 genotyped alleles and generated full-length gene sequence annotations for 9 previously incompletely annotated HLA class I allele sequences derived from 14 of the 300 genotyped alleles.CONCLUSIONS: The demonstrated utility of the Omixon Holotype HLA assay in combination with our downstream analytical framework to generate fully phased, full-length gene consensus sequences for the identification and characterization of novel HLA alleles, facilitates the study of HLA polymorphism beyond the antigen recognition domain in human health and disease.
- Open Access
- 31 August 2016
- Published by Frontiers Media SA in Frontiers in Genetics
- Vol. 7, 143
Post-transcriptional Regulation of BRCA2 through Interactions with miR-19a and miR-19b
Breast cancer type 2, early onset susceptibility gene (BRCA2) is a major component of the homologous recombination DNA repair pathway. It acts as a tumor suppressor whose function is often lost in cancers. Patients with specific mutations in the BRCA2 gene often display discrete clinical, histopathological and molecular features. However, a subset of sporadic cancers has wild type BRCA2 and display defects in the Homology-Directed Repair (HDR) pathway, which is the hallmark of ‘BRCAness’. The mechanisms by which BRCAness arises are not fully understood but post-transcriptional regulation of BRCA2 gene expression by microRNAs (miRNAs) may contribute to this phenotype. Here, we examine the post-transcriptional effects that some members of the six-miRNA cluster known as the miR-17/92 cluster have on the abundance of BRCA2’s messenger RNA (mRNA) and protein. We discuss two interactions involving the miR-19a and miR-19b members of the cluster and the 3´UTR of BRCA2’s mRNA. We investigated these miRNA:mRNA interactions in fifteen cell lines derived from pancreatic, breast, colon, and kidney tissue. We show that over-expression of these two miRNAs results in a concomitant decrease of BRCA2’s mRNA and protein expression in a subset of the tested cell lines. Additionally, using luciferase reporter assays we identified direct interactions between miR-19a/miR-19b and a miRNA response element (MRE) in BRCA2’s 3´UTR. Our results suggest that BRCA2 is subject to a complex post-transcriptional regulatory program that has specific dependencies on the genetic and phenotypic background of cell types.
- 31 August 2016
- Published by Frontiers Media SA in Frontiers in Genetics
- Vol. 7, 143
Open AccessPost-transcriptional Regulation of BRCA2 through Interactions with miR-19a and miR-19b
Breast cancer type 2, early onset susceptibility gene (BRCA2) is a major component of the homologous recombination DNA repair pathway. It acts as a tumor suppressor whose function is often lost in cancers. Patients with specific mutations in the BRCA2 gene often display discrete clinical, histopathological and molecular features. However, a subset of sporadic cancers has wild type BRCA2 and display defects in the Homology-Directed Repair (HDR) pathway, which is the hallmark of ‘BRCAness’. The mechanisms by which BRCAness arises are not fully understood but post-transcriptional regulation of BRCA2 gene expression by microRNAs (miRNAs) may contribute to this phenotype. Here, we examine the post-transcriptional effects that some members of the six-miRNA cluster known as the miR-17/92 cluster have on the abundance of BRCA2’s messenger RNA (mRNA) and protein. We discuss two interactions involving the miR-19a and miR-19b members of the cluster and the 3´UTR of BRCA2’s mRNA. We investigated these miRNA:mRNA interactions in fifteen cell lines derived from pancreatic, breast, colon, and kidney tissue. We show that over-expression of these two miRNAs results in a concomitant decrease of BRCA2’s mRNA and protein expression in a subset of the tested cell lines. Additionally, using luciferase reporter assays we identified direct interactions between miR-19a/miR-19b and a miRNA response element (MRE) in BRCA2’s 3´UTR. Our results suggest that BRCA2 is subject to a complex post-transcriptional regulatory program that has specific dependencies on the genetic and phenotypic background of cell types.
- Open Access
- 9 July 2016
- Published by Springer Nature in BMC Genomics
- Vol. 17 (1) , 1-14
The next generation of target capture technologies - large DNA fragment enrichment and sequencing determines regional genomic variation of high complexity
The ability to capture and sequence large contiguous DNA fragments represents a significant advancement towards the comprehensive characterization of complex genomic regions. While emerging sequencing platforms are capable of producing several kilobases-long reads, the fragment sizes generated by current DNA target enrichment technologies remain a limiting factor, producing DNA fragments generally shorter than 1 kbp. The DNA enrichment methodology described herein, Region-Specific Extraction (RSE), produces DNA segments in excess of 20 kbp in length. Coupling this enrichment method to appropriate sequencing platforms will significantly enhance the ability to generate complete and accurate sequence characterization of any genomic region without the need for reference-based assembly. RSE is a long-range DNA target capture methodology that relies on the specific hybridization of short (20-25 base) oligonucleotide primers to selected sequence motifs within the DNA target region. These capture primers are then enzymatically extended on the 3’-end, incorporating biotinylated nucleotides into the DNA. Streptavidin-coated beads are subsequently used to pull-down the original, long DNA template molecules via the newly synthesized, biotinylated DNA that is bound to them. We demonstrate the accuracy, simplicity and utility of the RSE method by capturing and sequencing a 4 Mbp stretch of the major histocompatibility complex (MHC). Our results show an average depth of coverage of 164X for the entire MHC. This depth of coverage contributes significantly to a 99.94 % total coverage of the targeted region and to an accuracy that is over 99.99 %. RSE represents a cost-effective target enrichment method capable of producing sequencing templates in excess of 20 kbp in length. The utility of our method has been proven to generate superior coverage across the MHC as compared to other commercially available methodologies, with the added advantage of producing longer sequencing templates amenable to DNA sequencing on recently developed platforms. Although our demonstration of the method does not utilize these DNA sequencing platforms directly, our results indicate that the capture of long DNA fragments produce superior coverage of the targeted region.
- 9 July 2016
- Published by Springer Nature in BMC Genomics
- Vol. 17 (1) , 1-14
Open AccessThe next generation of target capture technologies - large DNA fragment enrichment and sequencing determines regional genomic variation of high complexity
The ability to capture and sequence large contiguous DNA fragments represents a significant advancement towards the comprehensive characterization of complex genomic regions. While emerging sequencing platforms are capable of producing several kilobases-long reads, the fragment sizes generated by current DNA target enrichment technologies remain a limiting factor, producing DNA fragments generally shorter than 1 kbp. The DNA enrichment methodology described herein, Region-Specific Extraction (RSE), produces DNA segments in excess of 20 kbp in length. Coupling this enrichment method to appropriate sequencing platforms will significantly enhance the ability to generate complete and accurate sequence characterization of any genomic region without the need for reference-based assembly. RSE is a long-range DNA target capture methodology that relies on the specific hybridization of short (20-25 base) oligonucleotide primers to selected sequence motifs within the DNA target region. These capture primers are then enzymatically extended on the 3’-end, incorporating biotinylated nucleotides into the DNA. Streptavidin-coated beads are subsequently used to pull-down the original, long DNA template molecules via the newly synthesized, biotinylated DNA that is bound to them. We demonstrate the accuracy, simplicity and utility of the RSE method by capturing and sequencing a 4 Mbp stretch of the major histocompatibility complex (MHC). Our results show an average depth of coverage of 164X for the entire MHC. This depth of coverage contributes significantly to a 99.94 % total coverage of the targeted region and to an accuracy that is over 99.99 %. RSE represents a cost-effective target enrichment method capable of producing sequencing templates in excess of 20 kbp in length. The utility of our method has been proven to generate superior coverage across the MHC as compared to other commercially available methodologies, with the added advantage of producing longer sequencing templates amenable to DNA sequencing on recently developed platforms. Although our demonstration of the method does not utilize these DNA sequencing platforms directly, our results indicate that the capture of long DNA fragments produce superior coverage of the targeted region.
- 1 November 2015
- Published by Elsevier in Physica A: Statistical Mechanics and its Applications
- Vol. 438, 188-209
Measuring complexity, nonextensivity and chaos in the DNA sequence of the Major Histocompatibility Complex
- 1 November 2015
- Published by Elsevier in Physica A: Statistical Mechanics and its Applications
- Vol. 438, 188-209
Measuring complexity, nonextensivity and chaos in the DNA sequence of the Major Histocompatibility Complex