Search

Structure-based, virtual High-Throughput Screening (vHTS) methods for predicting ligand activity in drug discovery are important when there are no or relatively few known compounds that interact with a therapeutic target of interest. State-of-the-art computational vHTS necessarily relies on effective methods for pose sampling and docking and generating an accurate affinity score from the docked poses. However, proteins are dynamic; in vivo ligands bind to a conformational ensemble. In silico docking to the single conformation represented by a crystal structure can adversely affect the pose quality. Here, we introduce AtomNet PoseRanker (ANPR), a graph convolutional network trained to identify and rerank crystal-like ligand poses from a sampled ensemble of protein conformations and ligand poses. In contrast to conventional vHTS methods that incorporate receptor flexibility, a deep learning approach can internalize valid cognate and noncognate binding modes corresponding to distinct receptor conformations, thereby learning to infer and account for receptor flexibility even on single conformations. ANPR significantly enriched pose quality in docking to cognate and noncognate receptors of the PDBbind v2019 data set. Improved pose rankings that better represent experimentally observed ligand binding modes improve hit rates in vHTS campaigns and thereby advance computational drug discovery, especially for novel therapeutic targets or novel binding sites.

Structure-based, virtual High Throughput Screening (vHTS) methods for predicting ligand activity in drug discovery are important when there are no or relatively few known compounds that interact with a therapeutic target of interest. State-of-the-art computational vHTS necessarily relies on effective methods for pose sampling and docking to generate an accurate affinity score from the docked poses. However, proteins are dynamic; in vivo, ligands bind to a conformational ensemble. In silico docking to the single conformation represented by a crystal structure can adversely affect the pose quality. Here we introduce AtomNet PoseRanker, a graph convolutional network trained to identify, and re-rank crystal-like ligand poses from a sampled ensemble of protein conformations and ligand poses. In contrast to conventional vHTS methods that incorporate receptor flexibility, a deep learning approach can internalize valid cognate and non-cognate binding modes corresponding to distinct receptor conformations. AtomNet PoseRanker significantly enriched pose quality in docking to cognate and non-cognate receptors of the PDBbind v2019 dataset. Improved pose rankings that better represent experimentally observed ligand binding modes improve hit rates in vHTS campaigns, and thereby advance computational drug discovery, especially for novel therapeutic targets or novel binding sites.

We performed shallow single-cell sequencing of genomic DNA across 1475 cells from a cell-line, COLO829, to resolve overall complexity and clonality. This melanoma tumor-line has been previously characterized by multiple technologies and is a benchmark for evaluating somatic alterations. In some of these studies, COLO829 has shown conflicting and/or indeterminate copy number and, thus, single-cell sequencing provides a tool for gaining insight. Following shallow single-cell sequencing, we first identified at least four major sub-clones by discriminant analysis of principal components of single-cell copy number data. Based on clustering, break-point and loss of heterozygosity analysis of aggregated data from sub-clones, we identified distinct hallmark events that were validated within bulk sequencing and spectral karyotyping. In summary, COLO829 exhibits a classical Dutrillaux’s monosomic/trisomic pattern of karyotype evolution with endoreduplication, where consistent sub-clones emerge from the loss/gain of abnormal chromosomes. Overall, our results demonstrate how shallow copy number profiling can uncover hidden biological insights., Through shallow single-cell sequencing of genomic DNA followed by clustering analysis, Velazquez-Villarreal et al. reveal sub-clones of the melanoma cell line COLO829 and further identify and validate chromosome translocations and copy number changes. This study illustrates how copy number variation analysis can provide insights into cancer cell heterogeneity.

Sequencing the genomes of individual cancer cells provides the highest resolution of intratumoral heterogeneity. To enable high throughput single cell DNA-Seq across thousands of individual cells per sample, we developed a droplet-based, automated partitioning technology for whole genome sequencing. We applied this approach on a set of gastric cancer cell lines and a primary gastric tumor. In parallel, we conducted a separate single cell RNA-Seq analysis on these same cancers and used copy number to compare results. This joint study, covering thousands of single cell genomes and transcriptomes, revealed extensive cellular diversity based on distinct copy number changes, numerous subclonal populations and in the case of the primary tumor, subclonal gene expression signatures. We found genomic evidence of positive selection – where the percentage of replicating cells per clone is higher than expected – indicating ongoing tumor evolution. Our study demonstrates that joining single cell genomic DNA and transcriptomic features provides novel insights into cancer heterogeneity and biology.SIGNIFICANCEWe conducted a massively parallel DNA sequencing analysis on a set of gastric cancer cell lines and a primary gastric tumor in combination with a joint single cell RNA-Seq analysis. This joint study, covering thousands of single cell genomes and transcriptomes, revealed extensive cellular diversity based on distinct copy number changes, numerous subclonal populations and in the case of the primary tumor, subclonal gene expression signatures. We found genomic evidence of positive selection where the percentage of replicating cells per clone is higher than expected indicating ongoing tumor evolution. Our study demonstrates that combining single cell genomic DNA and transcriptomic features provides novel insights into cancer heterogeneity and biology.

The ability to sequence entire genomes at the level of single cells (SCs) is an essential tool for mapping heterogeneity in cancer. Knowledge of this variability provides insight into cancer dynamics and the efficacy of therapeutics. Copy number variant (CNV) analysis requires only sparse coverage creating an affordable means for measuring cellular differences genome-wide. However, current approaches to SC DNA sequencing suffer from low throughput, cumbersome workflows and remain accessible to only a few select laboratories. We have developed a microfluidic, partitioning-based solution that allows sensitive CNV detection of thousands of SCs concomitantly. Isolated cells or nuclei are treated to gain complete accessibility to genomic DNA then co-partitioned along with gel beads supporting oligonucleotide barcodes. Barcoded fragments, representing genomic abundance, are generated and converted into Illumina-compatible libraries. The workflow supports various sample types including primary cells, cultured cells, tissue-dissociated cells and nuclei extracted from flash-frozen tissues. Accompanying this platform we have also developed a scalable bioinformatics pipeline for genome alignment, normalization, copy-number detection, and clustering of SC profiles. Using clusters for CNV analysis we can reliably detect 100 kb scale events and, in large clones, identify CNV events down to tens of kilobases with high confidence. Analysis of our SC whole-genome sequencing data shows excellent coverage uniformity, comparable to leading plate-based SC amplification techniques. To validate the accuracy of CNV detection, calls are measured against orthogonal data from well-characterized cell lines including RPMI8226, HCC1954, and HCC1143 and high concordance is observed. Furthermore, we are able to observe multiple clones including CNVs in normal cell lines GM12878 and BJ suggesting absolute cell line validation requires whole-genome SC sequencing. Evaluating nuclei extracted from stage 1 breast carcinoma and stage 1 renal cell carcinoma we find clones exhibiting common somatic mutations. Furthermore, many cells exhibit extensive genomic heterogeneity, effects we attribute to DNA replication, with the frequency of such events elevated in fast growing samples such as tumors or fresh cell cultures. In conclusion, we demonstrate a novel droplet-based system that permits scalable CNV calling at the SC level enabling high resolution characterization of intra-tumor heterogeneity. It is our intention that a reliable and cost-effective system will encourage widespread adoption of SC analysis as a means of characterizing the progression of cancer, lead to more effective treatments and unveil genomic diversity in other genetically heterogenous systems such as neuronal cells, high-passage cell lines and samples from aging populations. Citation Format: Andrew Price, Jon Sorenson, Kamila Belhocine, Claudia Catalanotti, Zeljko Dzakula, Susana Jett, Viajy Kumar, Bill Lin, Tony Makarewicz, Alaina Puleo, Mohammad Rahimi, Sanjam Sawhney, Joe Shuga, Maengseok Song, Katie Sullivan-Bibee, Tobias Wheeler, Yifeng Yin, Michael Schnall-Levin, Rajiv Bharadwaj. A scalable microfluidic platform for determining cellular heterogeneity by copy number detection [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2018; 2018 Apr 14-18; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2018;78(13 Suppl):Abstract nr 3395.