Computational Deconvolution Research Articles

A New Computational Deconvolution Algorithm for the Analysis of Forensic DNA Mixtures with SNP Markers.

Single nucleotide polymorphisms (SNPs) support robust analysis on degraded DNA samples. However, the development of a systematic method to interpret the profiles derived from the mixtures is less studied, and it remains a challenge due to the bi-allelic nature of SNP markers. To improve the discriminating power of SNPs, this study explored bioinformatic strategies to analyze mixtures. Then, computer-generated mixtures were produced using real-world massively parallel sequencing (MPS) data from the single samples processed with the Precision ID Identity Panel. Moreover, the values of the frequency of major allele reads (FMAR) were calculated and applied as key parameters to deconvolve the two-person mixtures and estimate mixture ratios. Four custom R language scripts (three for autosomes and one for Y chromosome) were designed with the K-means clustering method as a core algorithm. Finally, the method was validated with real-world mixtures. The results indicated that the deconvolution accuracy for evenly balanced mixtures was 100% or close to 100%, which was the same as the deconvolution accuracy of inferring the genotypes of the major contributor of unevenly balanced mixtures. Meanwhile, the accuracy of inferring the genotypes of the minor contributor decreased as its proportion in the mixture decreased. Moreover, the estimated mixture ratio was almost equal to the actual ratio between 1:1 and 1:6. The method proposed in this study provides a new paradigm for mixture interpretation, especially for inferring contributor profiles of evenly balanced mixtures and the major contributor profile of unevenly balanced mixtures.

CUT&Tag2for1: a modified method for simultaneous profiling of the accessible and silenced regulome in single\xa0cells

Cleavage Under Targets and Tagmentation (CUT&Tag) is an antibody-directed transposase tethering strategy for in situ chromatin profiling in small samples and single cells. We describe a modified CUT&Tag protocol using a mixture of an antibody to the initiation form of RNA polymerase II (Pol2 Serine-5 phosphate) and an antibody to repressive Polycomb domains (H3K27me3) followed by computational signal deconvolution to produce high-resolution maps of both the active and repressive regulomes in single cells. The ability to seamlessly map active promoters, enhancers, and repressive regulatory elements using a single workflow provides a complete regulome profiling strategy suitable for high-throughput single-cell platforms.

DecOT: Bulk Deconvolution With Optimal Transport Loss Using a Single-Cell Reference.

Tissues are constituted of heterogeneous cell types. Although single-cell RNA sequencing has paved the way to a deeper understanding of organismal cellular composition, the high cost and technical noise have prevented its wide application. As an alternative, computational deconvolution of bulk tissues can be a cost-effective solution. In this study, we propose DecOT, a deconvolution method that uses the Wasserstein distance as a loss and applies scRNA-seq data as references to characterize the cell type composition from bulk tissue RNA-seq data. The Wasserstein loss in DecOT is able to utilize additional information from gene space. DecOT also applies an ensemble framework to integrate deconvolution results from multiple individuals’ references to mitigate the individual/batch effect. By benchmarking DecOT with four recently proposed square loss-based methods on pseudo-bulk data from four different single-cell data sets and real pancreatic islet bulk samples, we show that DecOT outperforms other methods and the ensemble framework is robust to the choice of references.

Reliability of a novel approach for reference-based cell type estimation in human placental DNA methylation studies

The placenta is a central organ during early development, influencing trajectories of health and disease. DNA methylation (DNAm) studies of human placenta improve our understanding of how its function relates to disease risk. However, DNAm studies can be biased by cell type heterogeneity, so it is essential to control for this in order to reduce confounding and increase precision. Computational cell type deconvolution approaches have proven to be very useful for this purpose. For human placenta, however, an assessment of the performance of these estimation methods is still lacking. Here, we examine the performance of a newly available reference-based cell type estimation approach and compare it to an often-used reference-free cell type estimation approach, namely RefFreeEWAS, in placental genome-wide DNAm samples taken at birth and from chorionic villus biopsies early in pregnancy using three independent studies comprising over 1000 samples. We found both reference-free and reference-based estimated cell type proportions to have predictive value for DNAm, however, reference-based cell type estimation outperformed reference-free estimation for the majority of data sets. Reference-based cell type estimations mirror previous histological knowledge on changes in cell type proportions through gestation. Further, CpGs whose variation in DNAm was largely explained by reference-based estimated cell type proportions were in the proximity of genes that are highly tissue-specific for placenta. This was not the case for reference-free estimated cell type proportions. We provide a list of these CpGs as a resource to help researchers to interpret results of existing studies and improve future DNAm studies of human placenta.

P032 Host transcriptome signatures in human fecal-washes predict histological remission in IBD patients

Abstract Background Colonoscopy is the gold standard for evaluation of inflammation in inflammatory bowel diseases (IBD), yet entails cumbersome preparations and risks of injury. Existing non-invasive prognostic tools are limited in their diagnostic power. Moreover, transcriptomics of colonic biopsies have been inconclusive in their association with clinical features. Our aim was to assess the utility of host transcriptomics of fecal wash samples of IBD patients compared to controls. Methods In this prospective cohort study, we obtained biopsies and fecal-wash samples from IBD patients and controls undergoing lower endoscopy. We performed RNAseq of biopsies and matching fecal-washes, and associated them with endoscopic and histological inflammation status. We also performed fecal mass-spectrometry proteomics on a subset of samples. We inferred cell compositions using computational deconvolution and used classification algorithms to identify informative genes. Results We analyzed biopsies and fecal washes from 39 patients (19 IBD, 20 controls). Host fecal-transcriptome carried information that was distinct from biopsy RNAseq and fecal proteomics. Transcriptomics of fecal washes, yet not of biopsies, from patients with histological inflammation were significantly correlated to one another (p=5.3*10–12). Fecal-transcriptome was significantly more powerful in identifying histological inflammation compared to transcriptome of intestinal biopsies (150 genes with area-under-the-curve >0.9 in fecal samples versus 10 genes in biopsy RNAseq). Fecal samples were enriched in inflammatory monocytes, regulatory T cells, natural killer-cells and innate lymphoid cells. Figure 1 - Fecal-wash host transcriptome predicts histological inflammation. A) Study layout, B) Clustergram of fecal-wash host cell mRNA signatures, demonstrating that patients with histological inflammation (red) are clustered when measuring fecal wash transcriptome yet not biopsy transcriptomes. C-D) Principle Component Analysis demonstrating improved separation of inflamed patients based on fecal host transcriptomes. E, F) Expression of host genes in fecal washes has higher statistical power (Area under the Curve, AUC) in classifying histological inflammation compared to biopsies. D shows NFKBIA as an example, E shows the AUC of the 5% best classifying genes, F shows the overall AUC based on biopsies or washes. Gray areas have AUC>0.9. G) UMAP of cells obtained from scRNAseq of mouse small intestine fecal washes. Conclusion Fecal wash host transcriptome is a powerful biomarker reflecting histological inflammation. Furthermore, it opens the way to identifying important correlates and therapeutic targets that may be obscure using biopsy transctriptomics.

Host transcriptome signatures in human faecal-washes predict histological remission in patients with IBD

BackgroundColonoscopy is the gold standard for evaluation of inflammation in inflammatory bowel diseases (IBDs), yet entails cumbersome preparations and risks of injury. Existing non-invasive prognostic tools are limited in their...

SwCAM: estimation of subtype-specific expressions in individual samples with unsupervised sample-wise deconvolution.

Complex biological tissues are often a heterogeneous mixture of several molecularly distinct cell subtypes. Both subtype compositions and subtype-specific (STS) expressions can vary across biological conditions. Computational deconvolution aims to dissect patterns of bulk tissue data into subtype compositions and STS expressions. Existing deconvolution methods can only estimate averaged STS expressions in a population, while many downstream analyses such as inferring co-expression networks in particular subtypes require subtype expression estimates in individual samples. However, individual-level deconvolution is a mathematically underdetermined problem because there are more variables than observations. We report a sample-wise Convex Analysis of Mixtures (swCAM) method that can estimate subtype proportions and STS expressions in individual samples from bulk tissue transcriptomes. We extend our previous CAM framework to include a new term accounting for between-sample variations and formulate swCAM as a nuclear-norm and ℓ2,1-norm regularized matrix factorization problem. We determine hyperparameter values using cross-validation with random entry exclusion and obtain a swCAM solution using an efficient alternating direction method of multipliers. Experimental results on realistic simulation data show that swCAM can accurately estimate STS expressions in individual samples and successfully extract co-expression networks in particular subtypes that are otherwise unobtainable using bulk data. In two real-world applications, swCAM analysis of bulk RNASeq data from brain tissue of cases and controls with bipolar disorder or Alzheimer's disease identified significant changes in cell proportion, expression pattern and co-expression module in patient neurons. Comparative evaluation of swCAM versus peer methods is also provided. The R Scripts of swCAM are freely available at https://github.com/Lululuella/swCAM. A user's guide and a vignette are provided. Supplementary data are available at Bioinformatics online.

Computational polarized Raman microscopy on sub-surface nanostructures with sub-diffraction-limit resolution.

Raman microscopy with resolution below the diffraction limit is demonstrated on sub-surface nanostructures. Unlike most other modalities for nanoscale measurements, our approach is able to image nanostructures buried several microns below the sample surface while still extracting details about the chemistry, strain, and temperature of the nanostructures. In this work, we demonstrate that combining polarized Raman microscopy adjusted to optimize edge enhancement effects and nanostructure contrast with fast computational deconvolution methods can improve the spatial resolution while preserving the flexibility of Raman microscopy. The cosine transform method demonstrated here enables significant computational speed-up from O(N3) to O(Nlog N) - resulting in computation times that are significantly below the image acquisition time. CMOS poly-Si nanostructures buried below 0.3 - 6 µm of complex dielectrics are used to quantify the performance of the instrument and the algorithm. The relative errors of the feature sizes, the relative chemical concentrations and the fill factors of the deconvoluted images are all approximately 10% compared with the ground truth. For the smallest poly-Si feature of 230 nm, the absolute error is approximately 25 nm.

UniDecCD: Deconvolution of Charge Detection-Mass Spectrometry Data.

Native mass spectrometry (MS) has become a versatile tool for characterizing high-mass complexes and measuring biomolecular interactions. Native MS usually requires the resolution of different charge states produced by electrospray ionization to measure the mass, which is difficult for highly heterogeneous samples that have overlapping and unresolvable charge states. Charge detection-mass spectrometry (CD-MS) seeks to address this challenge by simultaneously measuring the charge and m/z for isolated ions. However, CD-MS often shows uncertainty in the charge measurement that limits the resolution. To overcome this charge state uncertainty, we developed UniDecCD (UCD) software for computational deconvolution of CD-MS data, which significantly improves the resolution of CD-MS data. Here, we describe the UCD algorithm and demonstrate its ability to improve the CD-MS resolution of proteins, megadalton viral capsids, and heterogeneous nanodiscs made from natural lipid extracts. UCD provides a user-friendly interface that will increase the accessibility of CD-MS technology and provide a valuable new computational tool for CD-MS data analysis.

Dynamic non-line-of-sight imaging system based on the optimization of point spread functions.

Non-line-of-sight (NLOS) imaging reveals hidden objects reflected from diffusing surfaces or behind scattering media. NLOS reconstruction is usually achieved by computational deconvolution of time-resolved transient data from a scanning single-photon avalanche diode (SPAD) detection system. However, using such a system requires a lengthy acquisition, impossible for capturing dynamic NLOS scenes. We propose to use a novel SPAD array and an optimization-based computational method to achieve NLOS reconstruction of 20 frames per second (fps). The imaging system's high efficiency drastically reduces the acquisition time for each frame. The forward projection optimization method robustly reconstructs NLOS scenes from low SNR data collected by the SPAD array. Experiments were conducted over a wide range of dynamic scenes in comparison with confocal and phase-field methods. Under the same exposure time, the proposed algorithm shows superior performances among state-of-the-art methods. To better analyze and validate our system, we also used simulated scenes to validate the advantages through quantitative benchmarks such as PSNR, SSIM and total variation analysis. Our system is anticipated to have the potential to achieve video-rate NLOS imaging.

An unsupervised method for physical cell interaction profiling of complex tissues.

Cellular identity in complex multicellular organisms is determined in part by the physical organization of cells. However, large-scale investigation of the cellular interactome remains technically challenging. Here we develop cell interaction by multiplet sequencing (CIM-seq), an unsupervised and high-throughput method to analyze direct physical cell-cell interactions between cell types present in a tissue. CIM-seq is based on RNA sequencing of incompletely dissociated cells, followed by computational deconvolution into constituent cell types. CIM-seq estimates parameters such as number of cells and cell types in each multiplet directly from sequencing data, making it compatible with high-throughput droplet-based methods. When applied to gut epithelium or whole dissociated lung and spleen, CIM-seq correctly identifies known interactions, including those between different cell lineages and immune cells. In the colon, CIM-seq identifies a previously unrecognized goblet cell subtype expressing the wound-healing marker Plet1, which is directly adjacent to colonic stem cells. Our results demonstrate that CIM-seq is broadly applicable to unsupervised profiling of cell-type interactions in different tissue types.

Multiplatform computational analysis of mast cells in adrenocortical carcinoma tumor microenvironment

BackgroundImmunotherapeutic response failure of adrenocortical carcinomas highlights a need for novel strategies targeting immune cell populations in the tumor microenvironment to overcome tumor resistance and enhance therapeutic response. A recent study explored a new link between tumor mast cell infiltration and improved outcomes in patients with adrenocortical carcinomas. We further dissect the role of mast cells in the tumor microenvironment of adrenocortical carcinomas by examining the tumor mast cell expression signatures and mast cell activity within the tumor microenvironment to provide additional insight into potential novel immunotherapeutic targets. MethodsUsing the CIBERSORTx computational immunogenomic deconvolution algorithm to analyze adrenocortical carcinoma tumor gene messenger RNA expression data (The Cancer Genome Atlas, N = 79), we estimated the abundance of tumor immune infiltrating mast cells and assessed prognostic potential of mast cell signaling genes as pro or antitumor signatures, as well as examined the impact on overall and disease-free survival. ResultsWe stratified mast cell signaling genes with survival prognostic values (overall survival, disease-free survival, P < .05) into antitumor (ALOX5, CCL2, CCL5, CXCL10, HDC, IL16, TNF, TPSAB1, VEGFD) and protumor (CXCL1, CXCL3, CXCL8, IL4, IL13, PTGS3, TNSF4, VEGFD) groups. Antitumor mast cell signature, as the predominant phenotype, was associated with improved overall and disease-free survival. ConclusionThe deconvolution analysis of The Cancer Genome Atlas data identified mast cell infiltration in the adrenocortical carcinoma microenvironment as predominantly associated with antitumor activity. Future studies stemming from our findings may help define the role of mast cells in the tumor microenvironment and the impact on patient survival in patients with adrenocortical carcinomas. Modulation of tumor mast cell infiltration may serve as a potential target for novel synergistic immunotherapies for the treatment and improved survival of patients with adrenocortical carcinomas.

DNA methylation-based prediction of response to immune checkpoint inhibition in metastatic melanoma

BackgroundTherapies based on targeting immune checkpoints have revolutionized the treatment of metastatic melanoma in recent years. Still, biomarkers predicting long-term therapy responses are lacking.MethodsA novel approach of reference-free deconvolution of...

Molecular characterization of sequence-driven peptide glycation

Peptide glycation is an important, yet poorly understood reaction not only found in food but also in biological systems. The enormous heterogeneity of peptides and the complexity of glycation reactions impeded large-scale analysis of peptide derived glycation products and to understand both the contributing factors and how this affects the biological activity of peptides. Analyzing time-resolved Amadori product formation, we here explored site-specific glycation for 264 peptides. Intensity profiling together with in-depth computational sequence deconvolution resolved differences in peptide glycation based on microheterogeneity and revealed particularly reactive peptide collectives. These peptides feature potentially important sequence patterns that appear in several established bio- and sensory-active peptides from independent sources, which suggests that our approach serves system-wide applicability. We generated a pattern peptide map and propose that in peptide glycation the herein identified molecular checkpoints can be used as indication of sequence reactivity.

IMMU-14. COMPUTATIONAL DECONVOLUTION OF TUMOR-INFILTRATING IMMUNE COMPONENTS IN PEDIATRIC NERVOUS SYSTEM TUMORS

IntroductionIn the last decade, checkpoint inhibitor-based immunotherapy has been a groundbreaking development in the treatment of cancer. However, only a subset of patients treated with immune checkpoint inhibitors show long-lasting clinical benefit. Studies showed the tumor immune microenvironment (TME) as a particularly important factor influencing treatment response, critical for the design of other or combinatorial immunotherapy treatment strategies. Extensive research has been performed in the adult cancer field to unravel its immunogenomic aspects. However, in pediatric cancer this insight into tumor-infiltrating immune components is still lacking. This study aims to provide insight into the landscape of the immune microenvironment in pediatric primary nervous system tumors. MethodsBulk RNA-seq data of 936 pediatric primary solid tumors acquired from multiple international initiatives including Therapeutically Applicable Research To Generate Effective Treatments (TARGET), the International Cancer Genome Consortium (ICGC) and the Children’s Brain Tumor Tissue Consortium (CBTTC) were included in this study. We applied computational tumor immune microenvironment deconvolution, repurposed RNA-seq data to recover infiltrating T- and B-cell clonotypes and studied checkpoint gene expression across pediatric neural tumors. ResultsAmong pediatric neural tumors, embryonal tumors with multilayered rosettes (ETMR) and medulloblastomas (MB) were least immune infiltrated. Neuroblastomas (NBL) had the highest T-cell infiltration among pediatric cancers, while atypical teratoid/rhabdoid tumors (ATRT) had the highest levels of CD8 T cell infiltration among pediatric CNS tumors. While tumor mutational burden (TMB) was associated with immune cell infiltration in adult lung cancers and melanomas, we found no significant associations in pediatric cancers. The majority of NBL samples expressed LAG3, but ~10% of samples had elevated levels of TIM3 gene expression, suggesting a distinct mode of immunosuppression in this subset.

Microenvironment controls inflammatory response in macrophages

Abstract Inflammatory response often changes tissue environment, such as pH, oxygen level or composition of cells. Sustained deviation in these variables from homeostasis leads to tissue dysfunction and pathology, as collateral damage of inflammation. Meanwhile, we begin to realized that immune cells also constantly monitor tissue environment and maintain tissue homeostasis. This raises an intriguing question: do changes in tissue environment feedback to modulate immune functions during inflammation? pH is normally maintained within a narrow homeostatic range. Decrease in pH, or known as acidosis, is associated with diverse inflammatory conditions. Using macrophages, we found that acidic pH modulates their inflammatory response to lipopolysaccharide (LPS, TLR4 agonist) in a gene-specific manner, independent of known pH-sensing receptors. Employing a computational deconvolution algorithm, we systemically characterized how change in pH and activation of TLR4 regulate inflammatory gene expression. This revealed three distinct gene classes: a group insensitive to environment, and two groups representing alternative inflammatory programs, whose control by LPS is either antagonistic or synergistic with acidic pH. Surprisingly, this influence by pH correlates with differential control by NF-kB and IRF3, but is not due to difference in pathway activity. Through profiling chromatin landscape, we discovered that pH regulates gene expression via the activity of specific enhancers and chromatin remodeling. Our findings highlight a critical role of tissue microenvironment in controlling inflammatory gene expression, and reveal qualitatively different responses encoded by transcription circuits and epigenetic mechanisms.

Single-cell transcriptomics identifies gene expression networks driving differentiation and tumorigenesis in the human fallopian tube.

The human fallopian tube harbors the cell of origin for the majority of high-grade serous "ovarian" cancers (HGSCs), but its cellular composition, particularly the epithelial component, is poorly characterized. We perform single-cell transcriptomic profiling of around 53,000 individual cells from 12 primary fallopian specimens to map their major cell types. We identify 10 epithelial subpopulations with diverse transcriptional programs. Based on transcriptional signatures, we reconstruct a trajectory whereby secretory cells differentiate into ciliated cells via a RUNX3high intermediate. Computational deconvolution of advanced HGSCs identifies the "early secretory" population as a likely precursor state for the majority of HGSCs. Its signature comprises both epithelial and mesenchymal features and is enriched in mesenchymal-type HGSCs (p= 6.7× 10-27), a group known to have particularly poor prognoses. This cellular and molecular compendium of the human fallopian tube in cancer-free women is expected to advance our understanding of the earliest stages of fallopian epithelial neoplasia.

Abstract PO023: Single-cell transcriptomics identifies gene expression networks driving differentiation and tumorigenesis in the human fallopian tube

Abstract The human fallopian tube harbors the cell-of-origin for the majority of high-grade serous ‘ovarian’ cancers (HGSCs), but its cellular composition, particularly the epithelial component, is poorly characterized. We performed single-cell transcriptomic profiling in 12 primary fallopian specimens from 8 patients, analyzing around 53,000 individual cells to map the major immune, fibroblastic and epithelial cell types present in this organ. We identified 10 epithelial sub-populations, characterized by diverse transcriptional programs including SOX17 (enriched in secretory epithelial cells), TTF3 and RFX3 (enriched in ciliated cells) and NR2F2 (enriched in early, partially differentiated secretory cells). Based on transcriptional signatures, we reconstructed a trajectory whereby secretory cells differentiate into ciliated cells via a RUNX3high intermediate. Computational deconvolution of the cellular composition of advanced HGSCs based on epithelial subset signatures identified the ‘early secretory’ population as a likely precursor state for the majority of HGSCs. The signature of this rare population of cells comprised both epithelial (EPCAM, KRT) and mesenchymal (THY1, ACTA2) features, and was enriched in mesenchymal-type HGSCs, a group known to have particularly poor prognoses. This cellular and molecular compendium of the human fallopian tube in cancer-free women is expected to advance our understanding of the earliest stages of fallopian epithelial neoplasia. Further computational analyses on fibroblast and immune cells together with tumor single-cell data reveals extended contribution of non-epithelial microenvironment and cell-cell interactions and communications in fallopian tube that potentially drive HGSC progression. Citation Format: Huy Q. Dinh, Xianzhi Lin, Forough Abbasi, Robbin Nameki, Marcela Haro, Claire Olingy, Heidi Chang, Lourdes Hernandez, Simon Gayther, Kelly N. Wright, Paul-Joseph Aspuria, Beth Karlan, Rosario I. Corona, Andrew Li, B.J. Rimel, Matthew Siedhoff, Fabiola Medeiros, Kate Lawrenson. Single-cell transcriptomics identifies gene expression networks driving differentiation and tumorigenesis in the human fallopian tube [abstract]. In: Proceedings of the AACR Virtual Special Conference on the Evolving Tumor Microenvironment in Cancer Progression: Mechanisms and Emerging Therapeutic Opportunities; in association with the Tumor Microenvironment (TME) Working Group; 2021 Jan 11-12. Philadelphia (PA): AACR; Cancer Res 2021;81(5 Suppl):Abstract nr PO023.

Distinct Stromal and Immune Features Collectively Contribute to Long-Term Survival in Pancreatic Cancer

Background: The aggressive biology and treatment refractory nature of pancreatic ductal adenocarcinoma (PDAC) significantly limits long-term survival. Examining the tumor microenvironment (TME) of long-term survivors (LTS) of PDAC offers the potential of unveiling novel biological insights and therapeutic targets.Methods: We performed an integrated approach involving immunophenotyping, stromal scoring and histomorphological profiling of a cohort of 112 PDAC-cases, including 25 long-term survivors (LTSs, OS ≥ 60 months). Mutational frequencies were assessed using targeted next generation sequencing. Finally, we validated our findings in silico using an external cohort of microarray data from PDAC patients.Results: LTS cases exhibit a largely quiescent population of cancer-associated fibroblasts (CAFs). Immune profiling revealed key differences between LTS and NON-LTS cases in the intratumoral and stromal compartments. In both compartments, LTS cases exhibit a T cell inflamed profile with higher density of CD3+ T cells, CD4+ T cells, iNOS+ leukocytes and strikingly diminished numbers of CD68+ total macrophages, CD163+ (M2) macrophages and FOXP3+ Tregs. A large proportion of LTS cases exhibited tertiary lymphoid tissue (TLT) formation, which has been observed to be a positive prognostic marker in a number of tumor types. Using a Random-Forest variable selection approach, we identified the density of stromal iNOS+ cells and CD68+ cells as strong positive and negative prognostic variables, respectively. In an external cohort, computational cell-type deconvolution revealed a higher abundance of T cells, B lymphocytes and dendritic cells (DCs) in patients with long-term OS compared to short-term survivors. Thus, in silico profiling of long-term survivors in an external cohort, strongly corroborated the T cell-inflamed TME observed in our LTS group.Conclusions: Collectively, our findings highlight the prognostic importance of TME profiles in PDAC, underlining the crucial role of tumor associated macrophages (TAMs) and the potential interdependence between immunosuppressive TAMs and activated CAFs in pancreatic cancer. Additionally, our data has potential for precision medicine and patient stratification. Patients with a T cell inflamed TME might derive benefit from agonistic T cell antibodies (e.g., OX40 or CD137 agonists). Alternately, patients with activated CAFs and high infiltration of immunosuppressive TAMs are highly likely to exhibit therapeutic responses to macrophage targeted drugs (e.g., anti-CSF1R) and anti-CAF agents.

Abstract PS17-27: Molecular evaluation of immunogenicity and genomic alterations in invasive lobular breast cancer

Abstract Background: Invasive lobular carcinoma (ILC) is the second most common type of invasive breast cancer and accounts for 10-15% of all cases. Though ILC has distinct clinical, prognostic and molecular features, studies are limited and include few patients. ILCs show a decreased response to neoadjuvant chemotherapy and an increased resistance to endocrine therapy. Thus, there is a great need to identify alternative therapies, such as immunotherapy, that could improve overall survival. Success of immunotherapy largely depends on tumor immunogenicity which varies with histologic type. Determination of predictive and prognostic biomarkers for ILC will help determine who can benefit the most. Our study investigates canonical markers of immunogenicity - PD-L1 expression and Tumor Mutational Burden (TMB) - in patients with ILC compared to invasive ductal carcinoma (IDC). We further correlate these markers in different subtypes i.e. hormone receptor positive (HR+), HER2 positive (HER2+), and in triple negative breast cancers (TNBC). We also analyze differences in immune cell profiles constituting the tumor microenvironment (TME) and differences in genomic alterations between ILC and IDC. Methods: A retrospective data analysis was performed to identify breast cancer tumors with ILC or IDC histology profiled at Caris Life Sciences (Phoenix, AZ) that were tested for PD-L1 by SP142 assay and had whole transcriptome sequencing data available. ILC and IDC cases were further subtyped as HR+ and HER2+ and TNBC. PD-L1 expression in immune cell was assessed using Ventana PD-L1 (SP142) histochemical assay and PD-L1 expression in tumor cell was assessed by laboratory developed test using SP142 clone with staining higher than 2+ considered positive. TMB was measured by counting somatic non-synonymous missense mutations on the 592 gene panel (Nextseq) next generation sequencing (NGS) assay, and ≥ 10 mutations/megabase (mut/Mb) was considered high. Using the whole transcriptome RNA sequencing (NovaSeq) data we analyzed the difference in immune cell profiles constituting the TME using a computational RNA deconvolution approach. NGS was used to identify significant differences in genomic alterations between ILC and IDC. Results: We identified 1,359 breast cancer patients (ILC: 194 vs IDC: 1,165). Among ILC, 79% were HR +, 11% TNBC, &amp;lt;2% HER2+ and the rest were of unclear subtype, compared to 55% HR +, 31% TNBC, 10% HER2+ in the IDC group. PD-L1 expression in immune cells was lower in ILC (PD-L1≥ 1% ILC: 13% vs. IDC: 35% p &amp;lt;0.0001; PD-L1&amp;gt;10% ILC: 1% vs. IDC: 5% p=0.017). PD-L1 expression in tumor cells was also lower in ILC (ILC 1% vs IDC 5%; p =0.0136). All subtypes of ILC had lower PD-L1 expression in immune and tumor cells when compared to IDC. TMB was similar and comparable in IDC and ILC (median TMB 7 mut/Mb). Assessment of TME showed a significantly increased abundance of cytotoxic lymphocytes and monocytic cells in IDC, and stromal endothelial cells in ILC. Androgen receptor (AR) expression, and mutations in CDH1 (ILC 76% vs IDC 2%) and PIK3CA (ILC 44% vs IDC 30%) were more common in ILC. CDH1 mutation was significantly higher in both TN ILC (ILC 60% vs IDC 1%; p&amp;lt;0.05) and HR+ ILC (77% vs IDC 2%; p&amp;lt;0.05) subtypes of ILC. TP53 mutation was lower in ILC (ILC 26% vs IDC 65%) irrespective of subtype. Conclusion: In summary, PD-L1 expression in immune cells and tumor cells was lower in ILC, however TMB was comparable between ILC and IDC. Immune cell profiling supports a cold or less immunogenic TME for ILC. A composite immune biomarker may be able to better characterize immunogenicity of ILC. Citation Format: Upama Giri, Joanne Xiu, Tatiana Karadimitriou, Sofi Castanon, Foluso Ademuyiwa, Paula Pohlmann, Elias Obeid, Jasgit Sachdev, Elisa Krill-Jackson, Michael Simon, Antoinette Tan, Neelima Denduluri, Lee Schwartzberg, Michael Korn, Evanthia T Roussos Torres. Molecular evaluation of immunogenicity and genomic alterations in invasive lobular breast cancer [abstract]. In: Proceedings of the 2020 San Antonio Breast Cancer Virtual Symposium; 2020 Dec 8-11; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2021;81(4 Suppl):Abstract nr PS17-27.