DESIGNING LIGHTWEIGHT OPEN INNOVATION: A CONCEPTUALISATION OF HOW LARGE FIRMS ENGAGE WITH SMALL ENTREPRENEURIAL FIRMS

There is increasing scholarly interest in how large corporations engage in open innovation with small entrepreneurial firms, with synergies potentially producing positive outcomes for both the involved parties and the surrounding ecosystem. “Lightweight models” of open innovation (LOIs) have recently been introduced, governed by trust and relationships rather than by equity ownership and transactional control. This paper introduces a design framework and an alignment model for LOIs, based on 19 inductively generated and highly interrelated design elements associated with five design themes. The study uses empirical data from 18 LOI initiatives in Sweden, and the framework explains important differences in their motives, value propositions, innovation localizations, involved participants, and forms of interactions. Applying a value perspective to open innovation highlights two different value logics, suggesting that LOI initiatives can approach value by emphasizing either value creation or value capture. These logics may greatly influence other important design elements of LOIs.

Opportunism, bounded rationality and governance choices in exploration alliances: the moderating role of boundary spanners' guanxi

PurposeBy distinguishing opportunism-based and bounded rationality-based transaction costs, the study examines how firms use equity/relational governance and boundary spanners' guanxi to govern their exploration alliances in a transaction cost economizing way.Design/methodology/approachThis study used a survey methodology for data collection, and the sample consists of 150 exploration alliances formed by large Taiwanese information and electronic firms.FindingsFindings of this study show that exploration alliances incur considerable transaction costs and require high-level equity control and relational governance. The positive exploration of alliance-equity ownership relationship will be weakened by boundary spanners' guanxi when guanxi serves to harmonize conflicts and mitigate opportunism-based transaction costs, thereby reducing the need for using costly equity ownership to govern exploration alliances. In contrast, the positive exploration alliance-relational governance relationship will be amplified when guanxi becomes a source of legitimacy in the Chinese guanxi institution. This relation-augmenting effect will drive more relational governance because guanxi and relational governance together allow alliance managers to obtain sufficient legitimacy in the formation of a common dominant frame, thereby mitigating bounded rationality-based transaction costs.Originality/valueBy distinguishing various moderating effects of boundary spanners' guanxi and separating transaction costs into two forms, this study contributes to the existing literature as well as advances our understanding of alliance governance decisions in the Chinese business environment.

Evaluation of Functional and Cytogenetic High-Risk Multiple Myeloma Using Real-World Data

Introduction: Multiple myeloma (MM) is a heterogeneous disease with wide variability in outcomes. The presence of cytogenetic abnormalities in MM is of critical importance for prognosis and risk stratification. However, patients who may or may not have sufficient cytogenetic abnormalities to classify as high-risk can still experience rapid disease progression despite therapy, or functional high risk (FHR) disease. These two high risk cohorts comprise vulnerable subpopulations who have a significant burden of disease, and it is critical that we understand the underlying patient characteristics and optimal treatment sequence. We sought to investigate these two high risk patient populations treated in the contemporary real-world practice setting. Methods: A total of 1719 patients were identified in the COTA real-world database as having been diagnosed with active MM on or after January 1, 2015 and classified as either FHR, cytogenetic high risk (CHR), or both. The COTA real-world database is a USA-based real-world evidence database comprised of longitudinal, Health Insurance Portability and Accountability Act (HIPAA)-compliant, data on the diagnosis, clinical management, and outcomes of patients with cancer. Of the 1719 patients, 1260 were identified to be FHR, defined as relapse <18 months from initial active MM diagnosis. A total of 459 patients were identified as CHR, among which 347 were both FHR and CHR. CHR was defined as a patient having at least one of the following abnormalities: t(4;14), t(14;16), t(14;20), del(17p), 1q gain, or hypoploid. Line of therapy was applied programmatically using an algorithm based on International Myeloma Working Group criteria and clinical guidance. The primary outcome was time to next treatment (TTNT) calculated using the Kaplain Meier method. Univariate and multivariate analyses were conducted to understand predictors of rapid disease progression among high-risk patients. Results: In our real-world population, FHR patients tended to be slightly younger, African American, and treated predominantly in the academic setting (Table 1). First-line (1L) and second-line (2L) treatment patterns by category are shown in Table 2. A lower proportion of FHR patients received 1L immunomodulators as compared to the other high-risk groups, while almost half of the CHR patients received 1L stem cell transplant (SCT). In 2L, among patients not receiving 2L SCT, a higher proportion of CHR patients received a daratumamab-based treatment as compared to FHR (23.2% vs. 12.0%, respectively). We observed a longer median (95% CI) TTNT for high-risk patients receiving 2L daratumamab-based treatment as compared to patients who did not: 8.5 months (6.4-13.0) vs. 6.0 months (5.3-6.9), p=0.07 (Figure 1). Univariate and multivariate analyses showed age at diagnosis (HR: 0.98, CI: 0.97, 0.99), normal cytogenetics (HR: 0.78, CI: 0.63, 0.97), 1L immunomodulator (HR: 0.39, CI: 0.22, 0.69), 1L proteasome inhibitor (HR: 1.4, CI: 1.1, 1.8), and 1L SCT (HR: 0.22, CI: 0.15, 0.32) as significant predictors of rapid disease progression. Conclusions: Our study provides important insights comparing high risk populations with MM treated in the real-world setting. A higher proportion of CHR patients received 1L SCT and this provided the longest 1L TTNT as compared to other treatments. In 2L, among patients not receiving 2L SCT, we observed a trend towards significantly longer TTNT provided by dara-based treatment as compared to non-dara based treatment; however, our TTNT is lower than progression-free survival results observed in pivotal trials. We identified potential underlying differences in our patient populations that may be driving the predictors of rapid disease progression, including 1L SCT eligibility and renal disease, and these will be further investigated in propensity score matched populations. Future research will continue to explore optimal treatment sequences in high-risk populations with multiple myeloma to improve patient outcomes. These data highlight an urgent need to better predict FHR patients at diagnosis and develop clinical trials incorporating novel compounds in high risk patients. Figure 1 Figure 1. Disclosures Hansen: COTA, Inc.: Current Employment. Belli: COTA, Inc.: Current Employment, Other: Equity ownership. Goran: COTA, Inc.: Current Employment. Wang: COTA, Inc.: Current Employment, Other: Equity ownership.

Development of an Inclusive Risk Prognostic Index for Newly Diagnosed Multiple Myeloma

Background: Multiple myeloma (MM) risk stratification schemata such as the International Staging System (ISS) and Revised-ISS (R-ISS) were derived from clinical trial subjects made up predominately of younger White individuals with adequate renal function. It is unknown whether these prognostic indices are applicable to all patients with newly diagnosed (ND) MM, especially among Black individuals, older adults, and those with renal dysfunction. The R-ISS expanded on the ISS by including and serum lactate dehydrogenase (LDH) and high-risk cytogenetic abnormalities (HRCA) identified by fluorescence in-situ hybridization (FISH), but HRCA may not translate into poor prognosis for older adults and for Black individuals. We sought to create an inclusive risk prognostic index for NDMM using real-world data derived from electronic health records. Methods: De-identified NDMM patient-level data in the real-world setting was provided by COTA, Inc. 3000 patients were identified who met the inclusion criteria of NDMM between 2005 and 2020. Baseline diagnostic parameters available within 60 days before or after diagnosis were included. Progression free survival (PFS) was defined as the time from diagnosis to disease progression or death of any cause. Overall survival (OS) was defined as the time from diagnosis to death of any cause. Proportional hazards models were used to calculate hazard ratios (HR) and 95% confidence intervals for all-cause mortality. Age-adjusted univariate analyses identified variables significantly associated with OS, and continuous variables were dichotomized based on accepted cutoffs. Multivariate Cox models to identify the variables with the strongest association with OS were performed adjusting for age, sex, Black race, receipt of proteasome inhibitor and immunomodulatory imide during induction, autologous stem cell transplant within 1 year of diagnosis, ECOG performance status, and creatinine. An additive risk score was created with one point given to each significant variable. The risk score was then validated for PFS using the Multiple Myeloma Research Foundation's (MMRF) CoMMpass database (version IA15). Results: 3000 NDMM pts from the COTA, Inc. real-world database were initially evaluated, and a total of 689 NDMM pts had sufficient level of data to be included. The median follow-up time was 49.9 months (interquartile range (IQR) 29.1-76.2 months). Median age was 64 (IQR 32-86), including 44% age 65+. Of the 607 with reported race, 474 (78%) were White, 86 (14%) Black, 17 (3%) Asian, and 30 (5%) other. Of the 676 pts with reported serum creatinine (mg/dL), the median was 1 mg/dL (IQR 0.8-1.3) with 85 (13%) measuring >2 mg/dL. Examined peripheral blood variables were: calcium (corrected for albumin), albumin<3.5 mg/dL, beta2-microgloublin (B2M) >3.5 mcg/mL, LDH >250 U/L, hemoglobin <10 g/dL, M-spike >3 g/dL, and IgA isotype. Variables with significance using multivariate analysis at p<0.1 were: LDH>250 U/L, B2M >3.5 mcg/mL, hemoglobin <10 g/dL, and IgA isotype. These variables were simultaneously present in 558 patients. Patients were stratified into 3 groups: standard (std score = 0, n=186), intermediate (int score = 1-2, n=295), and high (score 3-4, n=77) risk. For this inclusive risk prognostic index (IRPI), the c-statistic was 0.61 for OS (HR 2.0, 95% CI 1.5-2.6, p<0.001) which compared favorably to the c-statistic for ISS (c=0.64, HR 1.8, 95% CI 1.5-2.2, p<0.001) and for R-ISS (c=0.63, HR 2.0, 95% CI 1.6-2.6). For the IRPI, median OS was 218 (std) vs 121.5 (int) vs 79.5 months (high). In comparison, median OS by ISS was 198.9 (stage I) vs 121.6 (stage II) vs 80.6 months (stage III), and by R-ISS: 198.9 (I) vs 121.6 (II) vs 79.5 months (III). Validation of the inclusive risk prognostic index (IRPI) using the MMRF CoMMpass database in 938 patients with all four criteria showed median PFS was 44 (std) vs 33 (int) vs 20.5 months (high). In comparison, median PFS by ISS was 45.9 (I) vs 31.5 (II) vs 20.5 (III) months. Median PFS by R-ISS was 50.1 (I) vs 32.7 (II) vs 19.1 (III) months. Conclusions: Employing real-world datasets that incorporate a more diverse patient population led to the generation of an inclusive risk prognostic index incorporating beta2-microgloublin, LDH, hemoglobin, and IgA isotype. This IRPI does not require bone marrow sampling, performs similarly to ISS and R-ISS in predicting PFS, and with datasets with longer follow-up may prove to predict OS. Figure 1 Figure 1. Disclosures Derman: Sanofi: Membership on an entity's Board of Directors or advisory committees. Belli: COTA, Inc.: Current Employment, Other: Equity ownership. Wang: COTA, Inc.: Current Employment, Other: Equity ownership. Hansen: COTA, Inc.: Current Employment. Jakubowiak: Amgen: Membership on an entity's Board of Directors or advisory committees; Karyopharm: Membership on an entity's Board of Directors or advisory committees; Gracell: Membership on an entity's Board of Directors or advisory committees; BMS: Membership on an entity's Board of Directors or advisory committees; GSK: Membership on an entity's Board of Directors or advisory committees; Abbvie: Membership on an entity's Board of Directors or advisory committees; Sanofi: Membership on an entity's Board of Directors or advisory committees; Janssen: Membership on an entity's Board of Directors or advisory committees; Celgene: Membership on an entity's Board of Directors or advisory committees.

Analyses of the Kinetics and Phenotype of Multiple Intraclonal CXCR4/CD5 B Cell Subsets Suggest Differences in Life Cycle Transitioning in CLL

Chronic lymphocytic leukemia (CLL) is a heterogeneous disease so that defining the dynamic features of the clone and its intraclonal subpopulations are essential to understand disease pathogenesis and to develop novel, effective therapies. For instance, because cell division is linked with new mutations, the ability to preferentially select cells that recently divided allows studying the subpopulation(s) most likely responsible for disease progression and resistance to therapies. The intraclonal kinetics of CLL B cells have been studied in clonal subgroups defined by reciprocal surface levels of CXCR4 and CD5. In that model, three fractions are identified: recently divided "proliferative" (PF; CXCR4 DimCD5 Bright); "intermediate" (IF; CXCR4 IntCD5 Int) and "resting" (RF; CXCR4 BrightCD5 Dim). Here, we have expanded the examination of subpopulations differing for time since last division ("age"). Unmanipulated CLL cells studied ex vivo from 10 patients who drank 2H 2O for 4 weeks were sorted by the relative densities of CXCR4 and CD5 to isolate the formerly identified PF, IF and RF as well as two fractions not previously characterized, "Double Dim" (DDF: CXCR4 DimCD5 Dim) and "Double Bright" (DBF; CXCR4 BrightCD5 Bright). For each fraction, the amount of deuterium incorporated into cellular DNA in vivo was measured. Consistently, the PF contained significantly higher levels of 2H-labeled DNA and higher calculated cell division rates when compared with the RF and IF. Interestingly, the DDF also contained significantly more 2H-labeled DNA compared to the RF; in contrast, the DBF resembled more closely the RF fraction. The overall 2H-incorporation gradient was: PF>DDF>IF>DBF>RF. In CLL, BCR signaling is fundamental, with the amount of membrane (m) IgM associating with signaling competence and disease aggressiveness. Additionally, when engaged independently, mIgM and mIgD can lead to different signaling sequelae. Therefore, we analyzed the 5 subpopulations for the densities of mIgM and mIgD. This showed a distribution similar to that of 2H-DNA incorporation: for IgM: PF=DDF>IF=DBF=RF, and for IgD: PF>DDF>IF=DBF>RF. Accordingly, we next measured 2H-DNA in subpopulations with low, intermediate and high levels of IgM and IgD. This revealed a direct correlation between IG densities and in vivo DNA synthesis, consistent with intraclonal subpopulations with high IGs having divided more recently than those with low IGs. However, these findings are not in line with cell division being primarily initiated by BCR engagement since that would lower mIgM levels. Therefore, we tested if engagement of TLR9 would affect mIG densities on CLL cells. After stimulation of 32 CLL clones with CpG+IL15, anti-IgM+IL4, anti-IgD+IL4, or anti-IgM-IgD+IL4, there was a significant increase in mIGs only after CpG+IL15 activation; each anti-IG stimulation led to downregulation of mIGs. Finally, we questioned the subclonal responsiveness to BTK inhibition in vivo. CLL samples taken from the same patients, before and during ibrutinib treatment, displayed intraclonal changes in mIG densities and cell size, the latter a marker of cellular and metabolic activation also linked with CLL in vivo birth rates. Ibrutinib treatment normalized mIgM and mIgD intraclonal densities and lead to an overall cell size decrease with larger, 2H-enriched and higher mIG density cells being more affected (PF>DDF>IF>DBF>RF). Collectively, these findings suggest that the most recently born cells enter the circulation as the PF from which they transition to either lower CD5 (DDF) or higher CXCR4 (IF and DBF) phenotypes. Each eventually converge as the RF. Moreover, since mIG densities on the more recently divided populations (PF and DDF) are high, the data imply that successful cell division is not solely a consequence of BCR engagement; the involvement of the TLR pathways, concomitantly or in series with BCR signaling, is more consistent with the higher mIG levels. Finally, ibrutinib treatment appears to preferentially target more recently divided cells with high mIG levels. Disclosures Allen: Alexion: Research Funding; Bristol Myers Squibb: Other: Equity Ownership; C4 Therapeutics: Other: Equity Ownership; Sanofi Genzyme: Membership on an entity's Board of Directors or advisory committees.

Serum Proteomic Analyses Suggest That the HMGB1 and Other Inflammatory Pathways Are Operational in MBL and Are Less in Overt CLL

Background. CLL-like monoclonal B-ceII lymphocytosis (MBL) is considered a requisite precursor of CLL, with 1-2% of subjects annually progressing to CLL requiring therapy. The role of immune alterations leading to and operating in MBL and controlling progression to CLL is not well characterized. Since an increased frequency of immune-related conditions associated with immune dysfunction exists prior to CLL diagnosis (Landgren et al, Br J Haematol 2007 and Blood 2007; Andersen et al, Leukemia 2020), immune alterations likely exist in MBL. We have examined the serum protein profiles of MBL and CLL in search of clues linking immune dysfunction with malignant transformation. Objectives. 1. Characterize the serum proteomic profiles of MBL individuals, IGHV-mutated CLL (M-CLL) and IGHV-unmutated CLL (U-CLL) patients, and age-matched healthy controls (HC). 2. Compare the serum proteomic profiles between the groups. 3. Assess the effect of IGHV mutation status on serum proteomic profiles in CLL. Methods. A total of 12 MBL, 12 M-CLL, 12 U-CLL and 12 age-matched HC were studied. Patients were cared for at Northwell Health, New York (n=25) and Hospital del Mar, Barcelona (n=11). All patients were treatment naive except for one U-CLL patient who was treated one year beforehand. Serum samples were collected, and their protein levels measured employing SOMAmers (modified single-stranded DNA aptamers; SomaLogic) to quantify relative levels of 1,310 proteins. P-values <0.05 were used to define significantly different protein levels between groups. Ingenuity Pathway Analysis (IPA; QIAGEN) was employed to evaluate potential protein interactions and pathways. IPA pathways with P-value <0.05 and Z-score ≥2 or ≤-2 were considered to be activated or inhibited, respectively. Results. Overall, the levels of 862 proteins differed significantly between groups: MBL vs. HC: 10 downregulated (d) and 24 upregulated (u); M-CLL vs. HC: 206d and 6u; U-CLL vs. HC: 54d and 40u; M-CLL vs. MBL: 384d and 5u; U-CLL vs. MBL: 74d and 24u; and M-CLL vs. U-CLL: 35d and 0u. IPA highlighted a role for the pro-inflammatory HMGB1 pathway in several comparisons. First, an activated HMGB1 pathway was predicted in MBL compared to HC, together with activation of phagocytes, and an inhibition of the systemic immune suppressor TGF-β. Second and consistent with the former, U-CLL patients displayed an activated HMGB1 pathway compared to HC, along with other signs of immune stimulation (activated NFkB and Th1 pathways, maturation of dendritic cells, and inhibition of TGF-β) and leukemic progression (activated progression of tumor, and leukocyte extravasation). Third and contrary to the above, the HMGB1 pathway was inhibited when comparing M-CLL to HC, in line with a global immune suppression signature (inhibited PRR, GM-CSF, FcεRI, IL1, IL8, TNF, Th1, STAT3 and NFkB pathways, in addition to inhibited cell movement, viability and activation). Notably, inhibition of immune pathways was predicted for both M-CLL and U-CLL compared to MBL (diminished TNF and IL6 signaling, and reduced cell movement), although the greatest differences were seen for M-CLL vs. MBL comparison, including blockade of the HMGB1 pathway in M-CLL. Finally, the M-CLL vs. U-CLL comparison suggested inhibited INFγ, IL2, IL3, IL4, NFkB, and decreased T lymphocyte stimulation and movement of tumor cells in M-CLL patients. Conclusions. An increased pro-inflammatory signature with involvement of the HMGB1 pathway was identified in MBL and U-CLL compared with HC, whereas the opposite was seen for M-CLL. Since MBL most often exhibit mutated IGHV (91% of cases in our cohort), these findings suggest immune stimulation as a characteristic feature in the pre-leukemic and U-CLL leukemic stage that surprisingly is not operative in M-CLL. Consistent with this, M-CLL displayed a global immune suppression (HMGB1 pathway inhibition), whereas U-CLL exhibited signs of immune stimulation (HMGB1 pathway activation) compared to HC which may relate to distinct capabilities of the two subtypes to interact with the microenvironment. Lastly, the increased inflammatory signature identified in MBL, which are mainly IGHV-mutated, was lessened in CLL, mainly in M-CLL patients. This is consistent with a decreased influence of immune imbalance and the HMGB1 pathway associated with IGHV-mutated clonal expansions. Acknowledgments. GB is supported by a grant from Fundación Alfonso Martín Escudero. Disclosures Allen: Alexon: Research Funding; Bristol Myers Squibb: Other: Equity Ownership; C4 Therapeutics: Other: Equity Ownership; Sanofi Genzyme: Membership on an entity's Board of Directors or advisory committees. Rhodes: Conquer Cancer Foundation Young Investigator Award: Other: Grant/Research Support; AbbVie, Genentech, Pharmacyclics, TG Therapeutics: Other: Consultant.

Therapeutic Targeting of Mertk and BCL-2 in T-Cell and Early T-Precursor Acute Lymphoblastic Leukemia

Introduction T-cell acute lymphoblastic leukemia (T-ALL) accounts for 15% of childhood ALL and is associated with inferior outcomes relative to B-cell ALL. Early T-precursor ALL (ETP-ALL) is a subset of T-ALL characterized by an immature T cell phenotype, resistance to therapy, and high rates of induction failure. MERTK receptor tyrosine kinase is ectopically expressed in 40-50% of T-ALLs, particularly those with an immature T cell phenotype, suggesting a role in ETP-ALL. Inhibition of MERTK using shRNA delayed leukemia progression and prolonged survival in a T-ALL xenograft model, implicating MERTK as a therapeutic target. MRX-2843 is an orally available dual MERTK/FLT3 inhibitor currently in phase I clinical trials. The anti-apoptotic protein B-cell lymphoma-2 (BCL-2) is specifically expressed in immature T cell precursors, is preferentially expressed in ETP-ALL compared to other T-ALLs, is essential for ETP-ALL cell survival, and is regulated downstream of MERTK in acute leukemia cells. Thus, combination therapies targeting these two proteins may be particularly effective to treat ETP-ALL. Methods Loucy and PEER ETP-ALL cell lines were cultured with vehicle or MRX-2843. Phosphorylated and total MERTK were assessed by immunoblot. Relative cell numbers were measured using Presto Blue reagent. Cells were stained with PoPro-1-iodide and propidium iodide and apoptotic and dead cells were quantitated by flow cytometry. T-ALL patient samples were cultured with UNC2025, a close analogue of MRX-2843, and relative cell numbers were assessed using MTS reagent. Orthotopic xenografts were established in NSG or NSGS mice using luciferase-expressing Jurkat cells (T-ALL), luciferase-expressing Loucy cells (ETP-ALL) or an ETP-ALL patient sample and leukemia burden was monitored by bioluminescence imaging or flow cytometry. MRX-2843 (65 mg/kg or 75 mg/kg) or saline vehicle were administered orally once daily. Differences in disease burden were assessed with the Mann-Whitney-U test or one-way ANOVA. Survival was determined by Kaplan-Meier analysis. Loucy and PEER cells were plated and screened in quadruplicate against >150 pairwise combinations of MRX-2843 and the BCL-2 inhibitor venetoclax in a high-throughput format. Synergy was calculated using the response additivity model. Results Treatment with MRX-2843 mediated a dose-dependent decrease in phosphorylated MERTK, inhibited expansion of ETP-ALL cells, and induced cell death in vitro. Fifty-four percent (21/39) of primary T-ALL patient samples were sensitive to UNC2025 with an IC 50≤550 nM, including 2/5 (40%) pediatric samples and 10/19 (53%) adolescent/young adult samples. Treatment with MRX-2843 significantly reduced leukemia burden in cell line-derived T-ALL and ETP-ALL xenograft models and prolonged survival by 50% and 13% in the T-ALL (n=10, p<0.0001) and ETP-ALL (n=10, p=0.0136) models, respectively. Similarly, in a patient-derived ETP-ALL xenograft model, treatment with MRX-2843 reduced peripheral blood disease burden by 83% and spleen weight by 64% compared to vehicle-treated mice (n=8, p<0.001) and prolonged survival by 41% (n=8, p=0.0016). MRX-2843 mediated anti-leukemia activity in combination with venetoclax and a dose ratio of 1:20 MRX-2843:venetoclax provided optimal synergy in Loucy and PEER ETP-ALL cells in vitro (Figure 1). Conclusions MRX-2843 has therapeutic activity in ETP-ALL cell culture and xenograft models and over half of T-ALL patient samples were sensitive to MERTK/FLT3 inhibition. MRX-2843 also mediated synergistic anti-leukemia activity against ETP-ALL cells in combination with venetoclax, with an optimal molar ratio of 1:20. These data demonstrate the therapeutic potential of MRX-2843 in patients with T-ALL, suggest that MRX-2843 may be particularly active alone and in combination with venetoclax in the ETP-ALL subset, and provide rationale for clinical testing of MRX-2843, with the ultimate goal to progress to trials evaluating MRX-2843 in combination with other agents. Toward this end, MRX-2843 monotherapy will be tested in patients with relapsed leukemia in an upcoming clinical trial (NCT04872478). Figure 1 Figure 1. Disclosures Wang: Meryx: Other: Equity ownership; University of North Carolina: Patents & Royalties. Frye: University of North Carolina: Patents & Royalties; Meryx: Membership on an entity's Board of Directors or advisory committees, Other: Equity ownership. Earp: Meryx: Membership on an entity's Board of Directors or advisory committees, Other: Equity ownership. Tyner: Petra: Research Funding; Incyte: Research Funding; Takeda: Research Funding; Janssen: Research Funding; Astrazeneca: Research Funding; Array: Research Funding; Constellation: Research Funding; Seattle Genetics: Research Funding; Schrodinger: Research Funding; Genentech: Research Funding; Gilead: Research Funding; Agios: Research Funding. DeRyckere: Meryx: Other: Equity ownership. Graham: Meryx: Membership on an entity's Board of Directors or advisory committees, Other: Equity ownership.

Pediatric Single Cell Cancer Atlas: An Integrative Web-Based Resource for Single Cell Transcriptome Data from Pediatric Leukemias

Introduction: The emergence and optimization of single cell profiling as a powerful tool to characterize the tumor microenvironment has revealed the heterogeneity of pediatric cancers, particularly different leukemia types/subtypes. Ease of access and analysis of the data from studies on different leukemia types is critical for improving diagnosis as well as therapy.Currently, there are no single cell data based resources available for pediatric leukemias. We have developed a comprehensive resource, Pediatric Single Cell Cancer Atlas (PedScAtlas), with the goal of developing a pan-leukemia genomics signature as well as highlighting the heterogeneity of different types of leukemia. This resource facilitates exploration and visualization of expression signatures in different leukemias without requiring extensive analysis and bioinformatics support. Methods: The PedScAtlas was built based on single cell data from various leukemias and normal bone marrow (BM) cells that have been pre-processed and analyzed using a uniform approach to generate normalized expression data (M. Bhasin et al. Blood 2020 (ASH), S. S. Bhasin et al. Blood 2020 (ASH), Panigraphy et al. JCI 2019, Stroopinsky et al. Haematologica 2021, Thomas et al. Blood 2020 (ASH)). The current version of PedScAtlas contains data from 30 local leukemia samples (Bhasin, et al. Blood 2020 (ASH), Thomas et al. Blood 2020 (ASH)) and those available on public resources such as GEO (Bailur et al. JCI Insight 2020). The PedScAtlas dataset and the Immune cell dataset each underwent quality control, integration, normalization, and dimensionality reduction using the Uniform Manifold Approximation and Projection (UMAP) method. Unsupervised UMAP analysis identified cellular clusters with similar transcriptome profiles that were annotated based on expression of cell-specific markers. Differential expression comparing different types of leukemia including acute myeloid leukemia (AML), B-cell acute lymphoblastic leukemia (B-ALL), T-cell acute lymphoblastic leukemia (T-ALL), and mixed phenotype acute leukemia (MPAL) samples was performed. To further ascertain genes specifically expressed in malignant blasts, the atlas also contains data from normal BM samples (Bailur et al. JCI Insight 2020) and healthy immune cells from the census of Immune Cells by the Human Cell Atlas Project (https://data.humancellatlas.org/). The web resource source code is written in R programming language and the interactive webserver has been implemented using the R Shiny package (Fig 1). The tool has been extensively tested on multiple operating systems (Linux, Mac, Windows) and web-browsers (Chrome, FireFox, and Safari). The tool is currently hosted on a 64bit CentOS 6 backend server running the Shiny Server program designed to host R Shiny applications. Results: The PedScAtlas contains data that facilitate exploration of gene expression profiles across leukemia types/subtypes and tumor microenvironment (TME) cell types. The atlas includes data from 33,930 AML, 25,744 T-ALL, 13,404 MPAL, and 6,252 B-ALL blast cells. It also contains single cell profiles of healthy BM samples from publicly available studies. The user can select data sets from the 5 major types of leukemia and normal BM in any combination of their choice to explore the expression profile of a gene of interest. The data can be visualized as UMAP (Fig. 1), or violin plots with annotations based on cluster ID, cell type, disease type, sample ID, and future continuous remission or relapse outcome. The UMAP with gene expression analyses allows the user to visualize the distribution of cell expressing a given gene on the UMAP plot. The Biomarker tool shows expression of different leukemia biomarker gene sets in the entire leukemia dataset. The Immune Cell section contains BM data from the Human Cell Atlas Project; the purpose of this section is to validate leukemia biomarkers by checking that the gene does not have significant expression in the healthy immune microenvironment. Conclusions: The PedScAtlas resource provides a unique and straightforward tool for biomarker identification, analysis of leukemia subtype heterogeneity, and transcriptome profile of the immune cell microenvironment. The resource is available online at https://bhasinlab.bmi.emory.edu/PediatricSC/. Figure 1 Figure 1. Disclosures DeRyckere: Meryx: Other: Equity ownership. Graham: Meryx: Membership on an entity's Board of Directors or advisory committees, Other: Equity ownership.