Search

Supplementary Figures S1-S2, Supplementary Methods, and Supplementary Tables S1-S8.

Purpose: Complete response to induction chemotherapy is observed in ∼60% of patients with newly diagnosed non-M3 acute myelogenous leukemia (AML). However, no methods exist to predict with high accuracy at the individual patient level the response to standard AML induction therapy.Experimental Design: We applied single-cell network profiling (SCNP) using flow cytometry, a tool that allows a comprehensive functional assessment of intracellular signaling pathways in heterogeneous tissues, to two training cohorts of AML samples (n = 34 and 88) to predict the likelihood of response to induction chemotherapy.Results: In the first study, univariate analysis identified multiple signaling “nodes” (readouts of modulated intracellular signaling proteins) that correlated with response (i.e., AUCROC ≥ 0.66; P ≤ 0.05) at a level greater than age. After accounting for age, similar findings were observed in the second study. For patients Conclusions: These data emphasize the value of performing quantitative SCNP under modulated conditions as a basis for the development of tests highly predictive for response to induction chemotherapy. SCNP provides information distinct from other known prognostic factors such as age, secondary AML, cytogenetics, and molecular alterations and is potentially combinable with the latter to improve clinical decision making. Independent validation studies are warranted. Clin Cancer Res; 16(14); 3721–33. ©2010 AACR.

Hepatic veno-occlusive disease/sinusoidal obstruction syndrome (VOD/SOS) is a potentially life-threatening complication of hematopoietic cell transplantation (HCT).To determine a blood biomarker signature early after HCT that identifies patients at high risk for VOD/SOS.A set of 23 plasma biomarkers, selected from the VOD/SOS literature, was measured on Days 0, 7, and 14 after myeloablative HCT using blood samples from patients enrolled in the Blood and Marrow Transplant Clinical Trials Network (BMT CTN) Protocol 1202. Eligible cases were diagnosed with VOD/SOS in BMT CTN 1202 using Baltimore criteria. Controls (without VOD/SOS) were matched to cases for conditioning regimen and age. Significant biomarkers were identified using Bonferroni-adjusted Wilcoxon rank sum (P≤0.002).33 patients with mild or severe VOD/SOS were identified (cases) and matched to 107 controls. Two, eight, and five biomarkers measured from the plasma of these patients were significantly associated with development of VOD/SOS at Days 0, 7, and 14, respectively, with the strongest associations being on Days 7 and 14. Biomarker associations were stronger for severe VOD/SOS risk and were stronger prognostic markers for VOD/SOS cases occurring within 28 days of HCT. Hyaluronan was most strongly associated with VOD/SOS risk, with an area under the receiving operator curve (AUC) of 0.81 on Day 7 and 0.79 on Day 14. Multivariate models of up to five biomarkers generated AUCs ranging from 0.82 to 0.85. All associations with VOD/SOS risk were independent of clinical risk factors.This study confirmed previously identified biomarkers of VOD/SOS risk and identified novel, prognostic biomarker signatures that identify patients at risk for VOD/SOS shortly after HCT. Multivariate analysis suggests that a combination of up to five of these protein biomarkers may provide a prognostic tool for identifying patients at risk for VOD/SOS.

Highlights • CTCs in NSCLC correlate with the presence of Treg cells expressing checkpoint markers in the tumor microenvironment. • Treg cells in NSCLC express CCR5, a new class of drug targets in inhibiting tumor growth and metastasis. • NSCLC tumors have a signature that is predictive of the presence of CTCs. • Highly multiplexed immune profiling reveals different patterns of expression in diploid compared to aneuploidy tumor cells from the same tumor. • Highly sensitive CTC enumeration methodology., Non-small cell lung cancer (NSCLC) has a poor prognosis. Targeted therapy and immunotherapy in recent years has significantly improved NSCLC patient outcome. In this study, we employed cell-by-cell immune and cancer marker profiling of the primary tumor cells to investigate possible signatures that might predict the presence or absence of circulating tumor cells (CTCs). We performed a comprehensive study on 10 NSCLC patient tissue samples with paired blood samples. The solid tissue biopsy samples were dissociated into single cells by non-enzymatic tissue homogenization and stained with a total 25 immune, cancer markers and DNA content dye and analyzed with high-parameter flow cytometry. CTCs were isolated and analyzed from the paired peripheral blood. We investigated a total of 74 biomarkers for their correlation with CTC number. Strong correlations were observed between CTC number and the frequency of immune checkpoint marker expressing lymphocytes (CTLA-4, LAG3, TIM3, PD-1), within the CD103+CD4+ T lymphocyte subset. CTC number is also correlated with the frequency of PD-L1 expressing cancer cells and cancer cell DNA content. In contrast, CTC number inversely correlated to the frequency of CD44+E-cadherin− cancer cells. Unsupervised clustering analysis based on the biomarker analysis separated the CTC negative patients from the CTC positive patients. Profiling multiple immune and cancer markers on cancer samples with multi-parametric flow cytometry allowed us to obtain protein expression information at the single cell level. Clustering analysis of the proteomic data revealed a signature driven by checkpoint marker expression on CD103+CD4+ T cells that could potentially be predictive of CTCs and targets of therapy.

Pierian Biosciences has developed the ChemoINTEL platform for high-throughput, semi-automated quantification of cellular responses to chemotherapeutics. ChemoINTEL relies on a microscopy-based process that captures and analyzes single cell behavior to reflect responses within a population. The platform is customizable to a variety of small molecule panels, either specific to certain tumor types or for titration screening of small molecules in development. Automated imaging of each well of a 384-well plate over time provides real-time kinetic response data under a variety of treatment conditions. Brightfield and fluorescent images are used to determine quantitative changes in morphology and molecular metrics reflecting induced apoptosis and cell death. Unique to ChemoINTEL as a drug screening platform is the quantification of intensity values, at the single cell level versus whole well fluorescence, for each fluorescent probe at each timepoint through a high-throughput analysis routine. These data are further processed through an internally developed algorithm to compare a treated population’s response relative to an untreated control and reports a sensitivity score. By combining different treatment conditions, the platform provides intelligent design of single agent or combination treatment approaches. By analyzing individual cells in the population, ChemoINTEL aids in better understanding response of tumor cell sub-populations to different chemotherapeutic agents. As a compound library screening tool or potential clinical diagnostic, Pierian Biosciences’ quality-controlled processes ensure all its equipment, reagents and processes follow ISO17025 guidelines to ensure the quality of all data generated. Each assay is internally controlled to include a standard cell line, whose response is monitored for accuracy of the reported results. In collaboration with several biorepositories and through the development of a standardized dissociation approach resulting in purification of viable primary tumor cells, the ChemoINTEL platform is also under development to assess chemotherapeutic response within patient populations. In conjunction with the ChemoINTEL platform, Pierian Biosciences has also developed a sophisticated multi-color flow cytometry platform (ImmunoINTEL) that when used in conjunction with ChemoINTEL provides information on the purity of isolated primary tumor cells prior to plating, as well as information on the cell populations present within the tumor microenvironment. These two platforms can be further complexed to evaluate immunotherapies in combination with front-line therapies. In summary, Pierian Biosciences has developed a high-throughput, multi-parametric platform for both drug development and diagnostic applications. Note: This abstract was not presented at the meeting. Citation Format: Kellye C. Kirkbride, Kevin J. Polach, Samantha J. Braxton, Megan D. Hoeksema, Dustin C. Rogers, Patricia Ladd-Ward, Durdica Vojnic Zelic, Santosh Putta, Matt Westfall, Norman Purvis. ChemoINTEL: A high-throughput, multi-parametric compound screening platform for intelligent lead compound and therapeutic combination identification [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2019; 2019 Mar 29-Apr 3; Atlanta, GA. Philadelphia (PA): AACR; Cancer Res 2019;79(13 Suppl):Abstract nr 294.

109 Background: Non-small cell lung cancer (NSCLC) has a poor prognosis as most patients are at advanced stage when diagnosed. Targeted therapy and immunotherapy in recent years has significantly improved NSCLC patient outcome. In this study, we employed cell-by-cell immune and cancer marker profiling of the primary tumor cells to investigate possible signatures that might predict the presence or absence of circulating tumor cells (CTCs). Methods: We performed a comprehensive study on 10 NSCLC patient tissue samples with paired blood samples. The solid tissue biopsy samples were dissociated into single cells by non-enzymatic tissue homogenization. The single cell suspensions were stained with a total 25 immune, cancer markers and a DNA content dye and analyzed with advanced, high-parameter flow cytometry. CTCs were isolated and analyzed from the paired peripheral blood. Results: Out of the 26 unique cell markers stained, we investigated a total of 72 biomarkers for their correlation with CTC number. Strong correlations were observed between CTC number and the frequency of immune checkpoint marker expressing lymphocytes, especially with the immune checkpoint marker expressing CD103+CD4+ T lymphocytes. CTC number is also correlated with the frequency of PD-L1 expressing cancer cells and cancer cell DNA content. In contrast, CTC number inversely correlated to the frequency of CD44+E-cadherin- cancer cells. Unsupervised clustering analysis based on the biomarker analysis separated the CTC negative patients from the CTC positive patients. Conclusions: Profiling multiple immune and cancer markers on cancer samples with multi-parametric flow cytometry allowed us to obtain protein expression information at the single cell level. Clustering analysis of the proteomic data revealed a signature driven by checkpoint marker expression on CD103+CD4+ T cells that could potentially be predictive of CTCs.

28 Background: Clinical benefits of immune checkpoint receptor (ICP-R) blockade in cancer patients can be predicted using various biomarkers such as Immune Index. However, traditional IHC approaches are lacking in sensitivity and in addressing the heterogeneity of the tumor microenvironment (TME), TIL functional status, expression of targetable ICP-R and their ligands (ICP-L), and detection of rare immune subsets within the TME. Methods: We developed a flow cytometry platform (ImmunoINTEL™) that delivers quantitative, clinically relevant information by merging these aspects of tumor biology into a single automated HTP assay comprised of multi-parameter Ab panels and automated sample/data processing procedures for immunophenotyping of blood and tissue specimens from patients with lung, breast, renal, and colorectal carcinomas. Results: Analysis of tumor specimens revealed distinct immune profiles characterized by variations in TIL makeup, differential expression of ICP-R and ICP-L (PD1/PDL1, TIGIT/CD112&CD155, Tim-3/Galectin-9, SIRPa/CD47), and changes in the functional/exhaustion status of TILs. We observed several discrete carcinoma profiles characterized by (1) an elevated Treg : CD8+GZMB+ T cell ratio, (2) high expression of ICP-L on epithelial cells and myeloid TILs, (3) enrichment for M1 macrophages (Mφ), M2 Mφ, and mixed M1/M2 phenotype Mφ (4) elevated levels of myeloid cells with suppressive phenotypes (MDSCs), (5) significant presence of myeloid and plasmacytoid dendritic cells, and (6) enrichment for mesenchymal cells. The elevated expression of several ICP-Ls on TILs and ICPs on epithelial cells suggested tumor- and TIL-intrinsic mechanisms modulating checkpoint interactions. Conclusions: The heterogeneity of TILs and ICP-R/ICP-L expression profiles across cell and tumor types can be explored in proof-of-mechanism studies and clinical trials to provide a rational for targeting distinct ICP-R/ICP-L interactions and TIL exhaustion mechanisms with immunotherapeutics, and for more refined selection of potential responders to ICP inhibitors.

Single cell network profiling (SCNP) is a multi-parameter flow cytometry technique for simultaneous interrogation of intracellular signalling pathways. Diagnostic paediatric acute myeloid leukaemia (AML) bone marrow samples were used to develop a classifier for response to induction therapy in 53 samples and validated in an independent set of 68 samples. The area under the curve of a receiver operating characteristic curve (AUC(ROC)) was calculated to be 0·85 in the training set and after exclusion of induction deaths, the AUC(ROC) of the classifier was 0·70 (P = 0·02) and 0·67 (P = 0·04) in the validation set when induction deaths (intent to treat) were included. The highest predictive accuracy was noted in the cytogenetic intermediate risk patients (AUC(ROC) 0·88, P = 0·002), a subgroup that lacks prognostic/predictive biomarkers for induction response. Only white blood cell count and cytogenetic risk were associated with response to induction therapy in the validation set. After controlling for these variables, the SCNP classifier score was associated with complete remission (P = 0·017), indicating that the classifier provides information independent of other clinical variables that were jointly associated with response. This is the first validation of an SCNP classifier to predict response to induction chemotherapy. Herein we demonstrate the usefulness of quantitative SCNP under modulated conditions to provide independent information on AML disease biology and induction response.

A greater understanding of the function of the human immune system at the single-cell level in healthy individuals is critical for discerning aberrant cellular behavior that occurs in settings such as autoimmunity, immunosenescence, and cancer. To achieve this goal, a systems-level approach capable of capturing the response of the interdependent immune cell types to external stimuli is required. In this study, an extensive characterization of signaling responses in multiple immune cell subpopulations within PBMCs from a cohort of 60 healthy donors was performed using single-cell network profiling (SCNP). SCNP is a multiparametric flow cytometry-based approach that enables the simultaneous measurement of basal and evoked signaling in multiple cell subsets within heterogeneous populations. In addition to establishing the interindividual degree of variation within a broad panel of immune signaling responses, the possible association of any observed variation with demographic variables including age and race was investigated. Using half of the donors as a training set, multiple age- and race-associated variations in signaling responses in discrete cell subsets were identified, and several were subsequently confirmed in the remaining samples (test set). Such associations may provide insight into age-related immune alterations associated with high infection rates and diminished protection following vaccination and into the basis for ethnic differences in autoimmune disease incidence and treatment response. SCNP allowed for the generation of a functional map of healthy immune cell signaling responses that can provide clinically relevant information regarding both the mechanisms underlying immune pathological conditions and the selection and effect of therapeutics.

A variety of machine learning algorithms, including hierarchical clustering, decision trees, k-nearest neighbours, support vector machines and bagging, were applied to construct models to predict the molecular weight of the polymers produced by a set of 96 homogeneous catalysts. The goal of the study was to develop models that could be used to screen large virtual libraries of catalysts in order to suggest candidates for further synthesis and screening. The descriptors used to represent the catalysts did not require detailed information about the catalysts themselves; they could be calculated using only the topology of the ligands. Using an initial set of five descriptors, model accuracies of about 70% were observed from each learning algorithm. A larger descriptor set (with ten descriptors) allowed bag classifiers that were 80% accurate to be built. All models were carefully evaluated to detect overfitting (memorization of the training data) and one example of the effects of overfitting is provided. Because the descriptors used in this study can be calculated very rapidly and the models themselves are very efficient, these bag classifiers are well suited to screening very large virtual libraries.

The performance of docking studies into protein active sites constructed by homology model building was investigated using CDK2 and factor VIIa screening data sets. When the sequence identity between model and template near the binding site area is greater than approximately 50%, roughly 5 times more active compounds are identified than would be found randomly. This performance is comparable to docking to crystal structures.

A study of the microstructure of polymers is essential for the understanding of their anomalous diffusion properties. The concept of free volume may be utilized to study relaxation phenomena and diffusion processes in polymers. Since a quantitative definition of free volume is not possible, a new concept called available volume is introduced. A simple geometric definition of available volume, relevant for the diffusion of small molecules in polymers, is given, based on the molecular structure. An initial molecular structure is obtained using a modified RIS (rotational isomeric state) approach. Then, with the aid of a molecular mechanics-based minimization procedure, computationally generated molecular structures for some polymers are used to estimate the available volume, using the proposed geometric definition. Numerically evaluated available volume distributions are then compared for different types of amorphous polymers, e.g. polycarbonates and polyacrylates, stressing their relevance to the diffusion properties.

Gemtuzumab ozogamicin (GO), an immunoconjugate between an anti-CD33 antibody and a calicheamicin-γ(1) derivative, induces remissions and improves survival in a subset of patients with acute myeloid leukemia (AML). As the mechanisms underlying GO and calicheamicin-γ(1) resistance are incompletely understood, we herein used flow cytometry-based single cell network profiling (SCNP) assays to study cellular responses of primary human AML cells to GO. Our data indicate that the extent of DNA damage is quantitatively impacted by CD33 expression and drug efflux activity. However, although DNA damage is required for GO-induced cytotoxicity, it is not sufficient for effective cell kill, suggesting that downstream anti-apoptotic pathways may function as relevant resistance mechanisms. Supporting this notion, we found activated PI3K/AKT signaling to be associated with GO resistance in vitro in primary AML cells. Consistently, the investigational AKT inhibitor MK-2206 significantly sensitized various human AML cells to GO or free calicheamicin-γ(1) with particularly pronounced effects in otherwise GO or free calicheamicin-γ(1)-resistant cells. Likewise, MK-2206 also sensitized primary AML cells to calicheamicin-γ(1). Together, our findings illustrate the capacity of SCNP assays to discover chemotherapy-related biological pathways and signaling networks relevant to GO-induced genotoxic stress. The identification of AKT signaling as being associated with GO resistance in vitro may provide a novel approach to improve the in vivo efficacy of GO/calicheamicin-γ(1) and, by extrapolation, other DNA damage-based therapeutics.

Background: Single cell network profiling (SCNP) is used to simultaneously measure the effects of modulators on signaling networks at the single cell level. SCNP-based biomarker assays predictive of response to induction therapy and relapse risk in acute myeloid leukemia (AML) patients are being developed. Such assays have typically used bone marrow (BM) as the sample source of blasts. Because circulating peripheral blasts are detectable in ∼65% of AML patients and peripheral blood (PB) sampling is less invasive than BM sampling, this study was performed to assess the effect of sample source on AML blasts signaling as measured in SCNP assay. Methods: SCNP using multiparametric flow cytometry was used to evaluate the activation state of intracellular signaling molecules in leukemic blasts under basal conditions and after treatment with modulators in 46 pairs of BM mononuclear cells/PB mononuclear cells. The relationship between readouts of modulated intracellular proteins (“nodes”) was measured using linear regression, Bland-Altman method, and Lin's concordance correlation coefficient. Results: The majority (156/161) of signaling nodes show strong correlations between paired PB and BM samples independently from the statistical method used. Notable exceptions were two PB samples with almost undetectable levels of circulating blasts compared with paired BM samples. Conclusions: Our results demonstrate that specimen source (BM or PB) does not significantly affect proteomic signaling in patients with AML and circulating blasts. The ability to use PB as a sample source will facilitate the monitoring of cellular signaling effects following administration of targeted therapies and at time points when BM aspirates are not clinically justifiable. © 2012 International Clinical Cytometry Society

BACKGROUND:Single cell network profiling (SCNP) utilizing flow cytometry measures alterations in intracellular signaling responses. Here SCNP was used to characterize Acute Myeloid Leukemia (AML) disease subtypes based on survival, DNA damage response and apoptosis pathways. METHODOLOGY AND PRINCIPAL FINDINGS:Thirty four diagnostic non-M3 AML samples from patients with known clinical outcome were treated with a panel of myeloid growth factors and cytokines, as well as with apoptosis-inducing agents. Analysis of induced Jak/Stat and PI3K pathway responses in blasts from individual patient samples identified subgroups with distinct signaling profiles that were not seen in the absence of a modulator. In vitro exposure of patient samples to etoposide, a DNA damaging agent, revealed three distinct "DNA damage response (DDR)/apoptosis" profiles: 1) AML blasts with a defective DDR and failure to undergo apoptosis; 2) AML blasts with proficient DDR and failure to undergo apoptosis; 3) AML blasts with proficiency in both DDR and apoptosis pathways. Notably, AML samples from clinical responders fell within the "DDR/apoptosis" proficient profile and, as well, had low PI3K and Jak/Stat signaling responses. In contrast, samples from clinical non responders had variable signaling profiles often with in vitro apoptotic failure and elevated PI3K pathway activity. Individual patient samples often harbored multiple, distinct, leukemia-associated cell populations identifiable by their surface marker expression, functional performance of signaling pathway in the face of cytokine or growth factor stimulation, as well as their response to apoptosis-inducing agents. CONCLUSIONS AND SIGNIFICANCE:Characterizing and tracking changes in intracellular pathway profiles in cell subpopulations both at baseline and under therapeutic pressure will likely have important clinical applications, potentially informing the selection of beneficial targeted agents, used either alone or in combination with chemotherapy.

Purpose: Complete response to induction chemotherapy is observed in ∼60% of patients with newly diagnosed non-M3 acute myelogenous leukemia (AML). However, no methods exist to predict with high accuracy at the individual patient level the response to standard AML induction therapy. Experimental Design: We applied single-cell network profiling (SCNP) using flow cytometry, a tool that allows a comprehensive functional assessment of intracellular signaling pathways in heterogeneous tissues, to two training cohorts of AML samples (n = 34 and 88) to predict the likelihood of response to induction chemotherapy. Results: In the first study, univariate analysis identified multiple signaling “nodes” (readouts of modulated intracellular signaling proteins) that correlated with response (i.e., AUCROC ≥ 0.66; P ≤ 0.05) at a level greater than age. After accounting for age, similar findings were observed in the second study. For patients Conclusions: These data emphasize the value of performing quantitative SCNP under modulated conditions as a basis for the development of tests highly predictive for response to induction chemotherapy. SCNP provides information distinct from other known prognostic factors such as age, secondary AML, cytogenetics, and molecular alterations and is potentially combinable with the latter to improve clinical decision making. Independent validation studies are warranted. Clin Cancer Res; 16(14); 3721–33. ©2010 AACR.

Measuring target coverage of small molecule inhibitors is paramount-first, for selection of molecules to progress through the drug development process and second, once a candidate drug moves to clinical testing, for guiding dose/schedule selection. Single cell network profiling (SCNP) using multiparameter flow cytometry can measure compound effects on multiple signaling cascades in a cell-type-specific manner. We applied SCNP to a panel of compounds with reported inhibitory effects on Jak/Stat signaling using a novel system where modulation of multiple signaling cascades are simultaneously measured in discrete cell subsets in whole (ie, unfractionated) blood. Jak2 vs. Jak3 selectivity as well as "off-target" effects on other cell signaling pathways were measured using a combination of cytokines that target different white blood cell subsets, namely GM-CSF (monocytes/granulocytes), IL-2 (T cells), and CD40L (B cells). The compounds were then rank-ordered by potency and selectivity against the different pathways tested. Notably, SCNP performed in whole unfractionated blood compared to fractionated peripheral blood mononuclear cells (PBMC) from the same donors revealed potency loss for all compounds, with one exception. These studies show that SCNP can be used to efficiently measure a drug candidate's potency and selectivity in a physiologically relevant environment (eg, whole blood) and that robust IC(50) are attainable from rare subpopulations (100 cells). The ability to generate in vitro IC(50) measurements in whole blood can be used not only for the preclinical selection of lead molecules, but also to determine the target plasma concentration for clinical studies and to measure target coverage after drug administration in early phase clinical trials. Knowledge of the compound plasma concentration necessary to achieve biochemical coverage permits rational design of clinical trials based on biologically active dose vs. the traditional maximum tolerated dose (MTD) design, which is better suited for cytotoxic, nontargeted drugs.

We review recent advances in computer modeling of molecular shape in drug discovery. We summarize the ways of representing shape computationally, discuss the various means of aligning molecules and shapes, consider the various ways of scoring similarity of shapes, and describe the ways in which these shapes can be used to construct molecular descriptors. Finally, we evaluate the success of these methods to date, suggest when they are best applied, and provide our recommendations for the direction of future work.

Discovering essential features shared by active compounds, an important step in drug-design, is complicated by conformational flexibility. We present a new algorithm to efficiently mine the conformational space of multiple actives and find small subsets of conformations likely to be biologically relevant. The approach identifies chemical and steric similarities between actives, providing insight into features important for binding when structural data are absent. Validation studies (thrombin and CDK2 data) produce alignments similar to protein-based alignments.

This review discusses the current challenges facing researchers developing computational models to predict absorption, distribution, metabolism, excretion and toxicity (ADMET) for early drug discovery. The strengths and weaknesses of different modeling approaches are reviewed and a survey of recent strategies to model several key ADMET parameters, including intestinal permeability, blood-brain barrier penetration, metabolism, bioavailability and drug toxicities, is presented.

We investigate the following data mining problem from computer-aided drug design: From a large collection of compounds, find those that bind to a target molecule in as few iterations of biochemical testing as possible. In each iteration a comparatively small batch of compounds is screened for binding activity toward this target. We employed the so-called “active learning paradigm” from Machine Learning for selecting the successive batches. Our main selection strategy is based on the maximum margin hyperplanegenerated by “Support Vector Machines”. This hyperplane separates the current set of active from the inactive compounds and has the largest possible distance from any labeled compound. We perform a thorough comparative study of various other selection strategies on data sets provided by DuPont Pharmaceuticals and show that the strategies based on the maximum margin hyperplane clearly outperform the simpler ones.

The shape of and the chemical features of a ligand are both critical for biological activity. This paper presents a strategy that uses these descriptors to build a computational model for virtual screening of bioactive compounds. Molecules are represented in a binary shape-feature descriptor space as bit-strings, and their relative activities are used to identify the subset of the bit-string that is most relevant to bioactivity. This subset is used to score virtual libraries. We describe the computational details of the method and present an example validation experiment on thrombin inhibitors.

A novel shape-feature-based computational method is described and used to rapidly filter compound libraries. The computational model, built using three-dimensional conformations of active and inactive molecules, consists of a collection of whole molecule shapes and chemical feature positions that are ranked according to their correlation with activity. A small ensemble of these shapes and features is used to filter virtual compound libraries. The method is applied to two thrombin data sets and is shown to be efficient in identifying novel scaffolds with enhanced hit rates.

Meeting abstracts Ipilimumab, an anti-CTLA-4 monoclonal antibody, is approved for treatment of unresectable/metastatic melanoma. Treatment benefits only a subset of patients (pts) and is associated with significant adverse effects. Biomarkers of pt response are needed. Single cell network profiling

Abstract 2489 Purpose: Standard Induction chemotherapy (Ara-C/daunorubicin, 3+7 regimen) in elderly patients (pts) with AML results in approximately 35–45% complete remission (CR) rate, and pts with resistant disease (RD) have a median survival of only 1–3 months. Developing a test for accurate prediction of response to standard induction therapy at the time of diagnosis may help inform treatment selection and improve clinical trial design. Methods: Single cell network profiling (SCNP) is a multi-parameter flow cytometry based approach for simultaneous interrogation of intracellular signaling pathways at a single cell level. SCNP was used to evaluate signaling profiles in leukemic blasts and to develop a classifier (DXSCNP) of response to induction therapy (CR/CRi, i=incomplete) in a Training Set of cryopreserved diagnostic samples (57 PB and 43 BM) collected from 74 non-M3 AML pts, aged 56+ treated with 3+7-based regimens on 4 SWOG clinical trials. SCNP intracellular readouts quantifying apoptotic response after 24 hrs in vitro treatment with Ara-C/Daunorubicin formed the inputs for DXSCNP. Pt and disease characteristics available either at diagnosis, i.e. relevant to induction therapy choices (e.g., age, WBC counts, FAB class, secondary AML, performance status – CLINICAL1), or available after start of induction therapy (CLINICAL2) were used to develop 2 clinical predictors (DXCLINICAL1 and DXCLINICAL2) of CR/CRi in the Training Set. The performance characteristics of these 3 classifiers were then tested separately in BM and PB sample sets. Specifically, classifier validation was performed in 2 independent BM sample sets (A: n=24 BM samples from ECOG E3999 trial; B: n=42 independent BM samples from the same 4 SWOG trials from which the Training samples were derived) and in 1 PB sample set (C: n=53, from the 4 SWOG trials; notably only 24 patients were shared between Set B and C)). The area under the receiver operating characteristic curve (AUROC) was used to measure each classifier's ability to predict response to 3+7 induction therapy. Out of bag estimates (OOB) of AUROC were calculated using the Training Set, and H0:AUROC=0.5 was tested against HA:AUROC>0.5 for each classifier in the Validation Sets. Results: As shown in Table 1, DXSCNP was a significant predictor of CR/CRi in BM samples. The AUROC for the DXSCNP classifier was 0.81 in the Training Set and 0.76 (p=0.01) and 0.72 (p=0.02) in Validation Sets A and B, respectively. No significant DXCLINICAL1 was identified from the 74 pts in the Training Set(OOB AUROC∼0.5). By contrast the AUROC of DXCLINICAL2 (which included inputs for cytogenetics, Flt3ITD, and NPM1 mutational status) was 0.63 (OOB) in the Training Set, and 0.61 (p=0.18) and 0.53 (p=0.38) in Validation Sets A and B, respectively. Of note, DXSCNP remained significant (p=0.03 and 0.04) when controlling for DXCLINICAL2 in both sample sets. Similar performance of DXSCNPwas observed in different pre-specified subgroups (although of small sample sizes) defined by pt, sample or disease characteristics. Using PB samples, the AUROCs for the DXSCNP classifier were, 0.87 (OOB) in the Training Set and 0.53 (p=ns) in the Validation Set C. However, the performance of DXSCNP in PB samples differed significantly between secondary vs. De novo AML i.e. AUROC= 0.24 vs 0.8 respectively. Of note, DXSCNPperformed similarly for the subset of de novo patients for whom paired BM with PB samples were tested (AUC=0.78 (PB) and 0.74 (BM)). Conclusion: This is the first study describing the successful validation of an SCNP-based classifier to predict response to standard induction chemotherapy in elderly AML pts with performance superior to clinical variables available at diagnosis. A similar SCNP classifier with overlapping cellular biology inputs was previously validated for pediatric AML. These results confirm the ability of quantitative SCNP readouts to provide independent and actionable information on disease biology and pt treatment choices. Independent validation in prospective studies is warranted. Disclosures: Cesano: Nodality, Inc: Employment, Equity Ownership. Gayko:Nodality, Inc.: Employment, Equity Ownership. Putta:Nodality, Inc.: Employment, Equity Ownership. Louie:Nodality, Inc.: Employment, Equity Ownership. Westfall:Nodality, Inc.: Employment, Equity Ownership. Purvis:Nodality, Inc.: Employment, Equity Ownership. Spellmeyer:Nodality, Inc.: Employment, Equity Ownership. Marimpietri:Nodality, Inc.: Employment, Equity Ownership. Hackett:Nodality, Inc.: Employment, Equity Ownership. Shi:Nodality, Inc.: Employment, Equity Ownership.

Abstract 3544 Background: In pediatric de novo AML, cytarabine (Ara-C)-based induction therapy results in above 80% complete response (CR) rates but nearly half of those who achieve an initial remission relapse die of their disease. Accurate prediction of initial response to chemotherapy at the time of diagnosis as well as identification of those at high risk of relapse despite initial remission would allow for patient specific therapy and improved clinical outcome. SCNP is a multiparametric flow cytometry-based assay that provides simultaneously quantitative measurements of extracellular surface markers as well as changes in intracellular signaling proteins in response to extracellular modulators at the single cell level (Kornblau et al. Clin Cancer Res 2010). In a previous study, we used this assay to define two distinct classifiers associated with response to standard induction therapy and risk of relapse in diagnostic bone marrow mononuclear cells from pediatric patients (pts) with non-M3 AML (ASH 2010;116:Abstract 954). This study is intended to confirm the validity of the pre-specified response to induction therapy classifier in an independent set of AML pediatric pts. Methods: The SCNP-based response classifier developed using 53 AML cryopreserved samples from patients enrolled on POG (now COG) trial 9421 was comprised of a combination of three SCNP readouts that measure apoptosis, MAPK signaling, and PI3K signaling and had a bootstrapped out-of-bag estimated Area Under the Receiver Operating Characteristic Curve (AUROC) of 0.84 (95% CI 0.67– 0.96). The classifier was tested on 68 cryopreserved samples (20 non–responders (NR) and 48 CRs) from patients enrolled on COG trials AAML0531 (samples from patients randomized to Ara-C, Daunomycin and Etoposide [ADE] induction therapy) and AAML03P1 (samples from patients treated with ADE plus Gemtuzumab Ozogamicin induction therapy). The primary hypothesis was that the prediction of induction response by the continuous score from the pre-specified classifier would yield an AUROC significantly greater than 0.5. Results: The primary objective of the study was met with an AUROC of 0.66 (n=68) p=0.042 (see table). The primary analysis used an NR classification that combined patients with either induction failure (n=14) or induction death (n=6). A pre-specified analysis in which induction deaths were removed resulted in an AUROC of 0.70 (n=62) p=0.021, suggesting that the underlying disease biology may be different for induction death vs. induction failure. In this study, WBC and cytogenetics risk groups were associated with induction response while age, gender and FLT3-ITD status were not. In a multivariate analysis of induction response that included WBC, cytogenetics and the pre-specified continuous SCNP classifier score, only cytogenetic risk group (p=0.001) and SCNP score (p=0.017) remained significant. Exploratory analyses excluding induction deaths suggest that the relationship between the SCNP score and induction response is strong among patients with an intermediate cytogenetic classification (n=23) (AUROC=0.88, p=0.002), while no relationship (AUC=0.48, p=0.959) is seen in those patients with a poor cytogenetic classification (n=17). Among the three SCNP signaling nodes contributing to the score, the node measuring drug-induced apoptosis performs most consistently across the training and validation sets. Conclusion: This study is the first validation of a SCNP-based classifier that predicts response to induction Rx. It shows that performing quantitative SCNP under modulated conditions can serve as the basis for developing biologically based tests in leukemia that offer new insights into the individual patient's disease biology. Additionally, this technology could prove useful in guiding alternative therapeutic strategies. Disclosures: Gayko: Nodality Inc.: Employment, Equity Ownership. Westfall:Nodality Inc.: Employment, Equity Ownership. Purvis:Nodality Inc.: Employment, Equity Ownership. Putta:Nodality Inc.: Employment, Equity Ownership. Hackett:Nodality Inc.: Employment, Equity Ownership. Cleary Cohen:Nodality Inc.: Consultancy, Equity Ownership.

Abstract 2512 Background: FMS-like tyrosine kinase 3 (FLT3) internal tandem duplication (ITD) mutations (FLT3 ITD+) result in constitutive activation of this receptor and have been shown to increase the risk of relapse in patients (pts) with AML; however, substantial heterogeneity in clinical outcomes still exists within both the FLT3 ITD+ and FLT3 ITD- AML subgroups, suggesting alternative mechanisms of disease relapse not accounted for by FLT3 mutational status. Single Cell Network Profiling (SCNP) is a multiparametric flow cytometry-based assay that simultaneously measures, in a quantitative fashion and at the single cell level, both extracellular surface marker levels and changes in intracellular signaling proteins in response to extracellular modulators (Kornblau et al. Clin Cancer Res 2010). Previously, we reported the use of this assay to functionally characterize FLT3 receptor signaling in healthy bone marrow and AML samples (Rosen et al. PLoS One 2010). By applying it to a separate cohort of samples collected from elderly non-M3 AML pts at diagnosis, a subclassification of AML samples beyond their “static” molecular FLT3 ITD status was generated (Rosen et al. ASH 2010 Abstr 2739). Specifically, FLT3 ITD- AML samples displayed a wide range of induced signaling, with a fraction having signaling profiles comparable to FLT3 ITD+ AML samples. Conversely, FLT3 ITD+ AML samples displayed more homogeneous induced signaling, with the exception of those with low mutational load, which had profiles more analogous to FLT3 ITD- AML samples. Due to the small numbers of pts in that exploratory study (n=44 [38 FLT3 ITD- and 6 FLT3 ITD+ pts]), an independent study was undertaken to confirm the observations, as well as to evaluate their clinical relevance (i.e., association with disease free survival (DFS) following anthracycline/cytarabine-based induction therapy). Methods: SCNP was performed as previously described on cryopreserved bone marrow or peripheral blood samples collected prior to anthracycline/cytarabine-based induction therapy from 104 elderly (>60y) non-M3 AML pts enrolled on ECOG trial 3999 or 3993 for whom ITD mutational status (including % mutational load), response and DFS data were available. Samples included 85 FLT3 ITD- and 19 FLT3 ITD+ AML, 30 and 8 of which, respectively, were collected from patients who achieved complete remission (CR). Objectives: The primary study objective was to confirm that levels of FLT3 ligand (FLT3L)-induced signaling (as measured by changes in intracellular phospho-S6 level) are more homogeneous in FLT3 ITD+ than in FLT3 ITD- myeloblasts. Four FLT3 ITD+ groups were pre-defined based on % mutation load (>0, 30%, 40%, or 50%). In addition, FLT3 ITD mutational status and signaling data from the SCNP assay (FLT3L and stem cell factor-induced phospho-S6 signaling and cytarabine/daunorubicin-induced apoptosis [cleaved PARP]) were combined to mathematically model their association with DFS among pts who achieved CR. DFS was defined as time from date of confirmed CR to date of relapse or death. Results: As shown in Figure 1a, our previous observations that variance in FLT3L-induced signaling is higher in FLT3 ITD- AML samples than in FLT3 ITD+ ones and that variance is decreased with increasing mutational load were verified in this study (Levene Test for FLT3 ITD- vs FLT3 ITD+ 50 p value=0.023). Further, when the association of DFS with FLT3 ITD mutational status and signaling data from the SCNP assay was measured using a Cox Proportional-Hazards model, the SCNP data were shown to provide independent information from FLT3 ITD mutational status (p =0.0115 for FLT3L-induced phospho-S6 signaling, Figure 1b). Conclusions: These data add to the growing body of evidence that, even within currently accepted risk stratification groups, AML is a heterogeneous disease. Functional characterization of FLT3 receptor signaling deregulation using SCNP provides prognostic information independent from FLT3 ITD mutational status and allows for more accurate pt stratification by functionally defining DFS risk sub-groups. Characterization of FLT3 signaling deregulation by SCNP could ultimately aid in the improved clinical management of AML pts and help identify candidates for FLT3 receptor inhibitor studies. Disclosures: Cesano: Nodality: Employment, Equity Ownership. Putta:Nodality Inc.: Employment, Equity Ownership. Mathi:Nodality: Employment. Rosen:Nodality Inc.: Employment, Equity Ownership. Gayko:Nodality Inc.: Employment, Equity Ownership. Hawtin:Nodality: Employment, Equity Ownership.

Abstract 954 Background: About 90% of pediatric patients with newly diagnosed acute myeloid leukemia (AML) have disease responsive to chemotherapy and achieve an initial complete response (CR) with anthracycline/cytarabine-based induction chemotherapy. However, 30–70% of patients of varying risks relapse within 4–5 years. Prospective identification of patients unlikely to benefit from induction therapy or likely to relapse could spare patients from treatment-related toxicities and allow consideration of alternative therapies. Current prognostic factors (e.g., cytogenetics and FLT3 ITD) of value at the population level are imperfect at the individual level. SCNP is a tool used to measure the effects of multiple modulators (including drugs) on signaling pathways at both the single-cell and individual patient level. A set of classifiers has been developed using SCNP technology that predicts the likelihood of response to anthracycline/cytarabine-based induction in adult patients with newly diagnosed AML. This training set has led to the development of classifiers that predict the response to anthracycline/cytarabine-based induction therapy and risk of disease relapse for pediatric patients with newly diagnosed AML. Validation in additional COG cohorts will lead to a novel SCNP classifier that can be used in the clinic. Methods: SCNP assays were performed on 77 bone marrow samples from pediatric AML patients enrolled in POG trial 9421 of which 67 were evaluable/had sufficient data for analysis and were enriched for non-responders (NR). 80 combinations of modulators and intra-cellular proteins (signaling nodes) were investigated including nodes involved in the phosphoinositide 3-kinase (PI3K), Janus Kinases (JAK), signal transducers and activators of transcription (STAT) and the DNA damage response and apoptosis pathways. Basal and modulated protein levels in leukemic blasts were measured using several metrics (e.g., fold change, total level of phosphorylation, and a rank based method Uu measuring the proportion of cells that shift from baseline), and nodes were examined in univariate and multivariate analyses for their ability to discriminate between responders (CR, n=46) and non-responders (NR, n=21) to anthracycline/cytarabine-based induction therapy. Furthermore, nodes were examined for their ability to identify patients likely to be in complete continuous remission (CCR, n=23) or relapse (CR-Rel, n=23) within 4 years. Results: Univariate analysis revealed 19 nodes associated with disease response to conventional induction therapy and 9 associated with CR-Rel (i.e., area under the operator/receiver curve (AUC of ROC=AUC*) >0.65; p0.8). Additional analyses that evaluate independence and ability to combine clinical or molecular predictors (e.g., cytogenetics, FLT3 ITD) with SCNP data will be presented. Conclusion: The training study data show the value of performing quantitative SCNP under modulated conditions as the basis for developing highly predictive tests for response to induction chemotherapy in pediatric patients with newly diagnosed AML. Independent validation studies will follow. Disclosures: Lacayo: Nodality Inc.: Honoraria. Cohen:Nodality Inc.: Employment, Equity Ownership. Westfall:Nodality Inc.: Employment, Equity Ownership. Lackey:Nodality Inc.: Employment, Equity Ownership. Xin:Nodality Inc.: Employment, Equity Ownership. Gayko:Nodality Inc.: Employment, Equity Ownership. Putta:Nodality Inc.: Employment, Equity Ownership. Cesano:Nodality Inc.: Employment, Equity Ownership.

Abstract 4843 Background: All major molecular technology platforms have used cryopreserved samples as a source in biomarker development studies. Such samples enable the conduct of prospectively designed studies using banked samples with relevant clinical annotations, thus potentially reducing the duration of biomarker development. Single Cell Network Profiling (SCNP) is a new technology platform requiring viable cells. It uses multi-parametric flow cytometry to measure biological pathways focusing on intra-cellular signaling and post-translational modulations in response to a variety of modulators (growth factors, cytokines, drugs, etc.). As part of a collaboration between industry, academic institutions, and cooperative groups to improve acute myeloid leukemia (AML) management, we are developing clinical tests based on SCNP to predict induction chemotherapy response or risk of relapse. To date, all the development studies have been performed on DMSO cryopreserved specimens of ficoll fractionated bone marrow (BM) or peripheral blood (PB) mononuclear cells. Demonstrating a correlation of SCNP readouts between fresh and cryopreserved preparations is essential to applying these proteomic signatures to the clinical setting. Objectives: To compare the results of SCNP assays between paired fresh and cryopreserved samples in a multicenter prospective study. Methods: To date, 13 fresh BM and PB samples were prospectively collected from pediatric or adult non-M3 AML patients at 3 academic centers and shipped over night to the Nodality CLIA lab. Samples were required to have 2 million viable cells per aliquot for SCNP assays, and underwent ficoll separation and mononuclear cells were divided into 2 aliquots - one processed fresh, and the second cryopreserved for 1 month, and then thawed and processed for the SCNP assay. 70 SCNP node-metrics (i.e. proteomic readouts in the presence or absence of modulator), identified previously as candidate proteomic signatures for several assays in development (including PIK3, Jak/STAT and DNA damage/apoptosis pathways) were investigated. The assay readouts for blast cells from a fresh aliquot were compared to the results from a cryopreserved aliquot by linear regression, Bland-Altman, and Lin's concordance analysis. Results: The analysis of paired aliquots from 13 patients, with median WBC of 27.9 (3-60) 10e3/ul, showed that cryopreservation did not affect sample quality as measured by percent of cells that were negative for cleaved PARP expression (R2 = 0.92 cryopreserved vs. fresh). The majority of unmodulated node-metrics (59%) and modulated node-metrics (68%, see Table) had a good correlation between the two preparations as measured by linear regression i.e., R2 > 0.64. The node-metrics with a lower R2 were using either a dim fluorophore (i.e. Alexa-647) and/or were within the low signal range (e.g., Erk basal); and therefore were not good candidates for future test development. Results using both Bland Altman and Lin's Concordance methods showed good concordance. Assessment of aliquots with up to 4 years in cryopreservation is ongoing and data from the 6 month time-point will be presented. Conclusions: These studies highlight the importance of cryopreservation of AML samples at clinical sites and by cooperative groups. These results demonstrate that cryopreservation maintains the activation signaling potential of AML blasts. SCNP assays developed and validated using cryopreserved samples can be applied to fresh samples and integrated prospectively into frontline clinical trials and clinical practice. Disclosures: Cesano: Nodality Inc.: Employment, Equity Ownership. Gotlib: Nodality Inc.: Honoraria. Lacayo: Nodality Inc.: Honoraria. Putta: Nodality Inc.: Employment, Equity Ownership. Lackey: Nodality Inc.: Employment, Equity Ownership. Gayko: Nodality Inc.: Employment, Equity Ownership. Kornblau: Nodality Inc.: Honoraria.

Abstract 2739 Background: Internal tandem duplication mutations of the FMS-like tyrosine kinase 3 receptor (FLT3-ITD) are among the most frequent mutations in de novo acute myeloblastic leukemia (AML) and result in constitutive activation of the receptor tyrosine kinase. Based on data showing that these mutations negatively affect outcome in AML, FLT3 receptor kinase activity is currently being targeted in the clinic. To date, only FLT3 receptor mutant AML patients have been selected for trials involving FLT3 receptor kinase inhibitors, with surprisingly variable clinical responses. Downstream targets of FLT3 receptor activation, whether by receptor mutation or FLT3 ligand binding, involve Stat5, PI3-kinase/Akt and the Ras/Raf/Erk kinase signal transduction pathways which ultimately affect cell survival and proliferation. Functional characterization of those signaling pathways in mutated versus non-mutated FLT3 receptor (FLT3 -WT) myeloblasts has significant potential to reveal heterogeneity among these genetically defined subgroups and to predict for response to kinase inhibition, independent of FLT3 receptor mutation status. Methods: Modulated single cell network profiling (SCNP) was used to evaluate the activation state of intracellular signaling molecules (i.e. nodes), including phosphorylated (p)-Akt, p-Erk, p-S6, p-Stat5 and cleaved-PARP, at baseline and after treatment with specific modulators [including cytokines (such as IL-27) growth factors (such as FLT3 ligand) and drugs (such as cytosine arabinoside)] in 7 healthy bone marrow mononuclear blasts (BMMb) and leukemic myeloblasts, characterized for FLT3 receptor mutation status, from 44 AML patients (38 FLT-WT and 6 FLT3-ITD), aged >60 years (ECOG trial E3999). A total of 64 node-metrics were analyzed. Results: Signaling profiles differed significantly in FLT3-ITD vs. FLT3-WT AML, and in FLT3-WT vs. BMMb (shown in Figure 1 for a representative node, FLT3 ligand induced p-S6). Specifically, compared to BMMb, FLT3-ITD blasts uniformly showed increased basal p-Stat5 levels, decreased FLT3 ligand-induced activation of P13K and Raf/Ras/Erk pathways, minimal IL-27 induced activation of the Jak/Stat pathway, and higher apoptotic responses to DNA-damaging agents. Two AML samples harboring a low FLT3-ITD mutant burden, however, exhibited a signaling pattern similar to FLT3-WT AML samples. By contrast, FLT3-WT samples displayed heterogeneous signaling profiles, overlapping both with those of FLT3-ITD and BMMb samples, suggesting that a fraction of FLT3-WT AML samples exhibit FLT3 receptor pathway deregulation even without FLT3-ITD. Conclusions This study showed that SCNP, which provides a detailed view of intracellular signaling networks at the single-cell level, subclassified patients with AML beyond their molecularly determined FLT3 mutation status. In particular, a fraction of FLT3-WT AML samples signaled as if containing a FLT3 receptor length mutation while FLT3-ITD samples with low mutational load signaled like FLT3-WT AML samples. The clinical relevance of this observation, both for disease prognosis and response to kinase inhibitors, will be revealed only if AML patients are accrued to kinase inhibition trials irrespective of FLT3 receptor mutation status. The wide range of signaling responses observed in FLT3-WT AML suggests that disease across FLT3-WT patients is heterogeneous, likely promoted through distinct mutations and alterations, giving rise to distinct signaling profiles in individual patients Our data also provide evidence for the co-existence of differentially signaling blast populations in individual patients. The potential impact of signaling heterogeneity on clinical response needs to be assessed and may require an individualized combination of treatment modalities. Disclosure: Rosen: Nodality Inc.: Employment, Equity Ownership. Putta:Nodality Inc.: Employment, Equity Ownership. Gayko:Nodality Inc.: Employment, Equity Ownership. Cesano:Nodality Inc.: Employment, Equity Ownership.

Abstract 2695 Background: Age and cytogenetics are major determinants of outcome in older patients with AML. Complete response (CR) to induction chemotherapy is observed in less than half of de novo AML patients (excluding APL) older than 60 years. Because unfavorable treatment results are often obtained despite intensive chemotherapy, which carries the potential for significant treatment-induced morbidity and mortality, it would be valuable to be able to predict with high accuracy at the individual patient level who will achieve a CR. Prospective identification of patients unlikely to benefit from anthracycline/cytarabine-based chemotherapy could spare patients from unnecessary treatment-related toxicities and allow consideration for alternative therapeutic strategies. Single Cell Network Profiling (SCNP) is used to measure simultaneously the effects of multiple modulators (including drugs) on signaling cascades at the single cell level. Using SCNP technology we have developed a set of classifiers that predict for likelihood of response to anthracycline/cytarabine-based induction therapy in older patients with AML. Methods: SCNP assays were performed on paired, bone marrow (BM) and peripheral blood (PB), samples from 44 AML patients (de novo, evolved from an antecedent MDS or MPN or treatment related), >60 years old, enrolled on ECOG trial E3999. Based on two previous training studies, 38 combinations of modulators and intra-cellular proteins (signaling nodes along the phosphoinositide 3-kinase (P13K ), the Janus Kinases (Jak) signal transducers and activators of transcription (Stat) and the DNA damage response and apoptosis pathways) were investigated. Basal and modulated protein levels and the effect of modulation on proteins levels in the leukemic blast cells were expressed using a variety of metrics. A total of 64 node/metric combinations (dimensions) were used to build multi-parametric classifiers (ranging from 2 to 10 nodes/metrics) using different modeling methodologies (including random forest, boosting, lasso and a bootstrapped best subsets logistic modeling approach that shrinks regression coefficients (BBLRS)) able to predict the likelihood of response to induction therapy. The performance characteristics of the classifiers built on the BM samples were then evaluated independently on the paired PB samples. Results: Several promising models with high area under the operator/receiver curve (AUROC) values (indicating strong agreement between actual clinical responses and responses as predicted by the model) were developed based on SCNP proteomic read outs for BM samples. The observed and predicted values from the current best BBLRS model are shown in Figure 1. The unadjusted AUROC of this model is 0.98 and the expected AUROC for the model when applied to an independent (validation) sample is 0.84. Five signaling nodes are represented in this model; they include nodes belonging to growth factor- induced survival pathways (PI3K, RAS/MAPK) as well as DNA damage response and apoptosis pathways. When the predictive accuracy of the lead SCNP classifier was compared to that of a model based on traditional clinical/molecular predictors (i.e. the combination of age, therapy-related AML, and karyotype) the adjusted AUROC of the SCNP classifier far surpassed that of the clinical predictors (adjusted AUROC=0.61 for clinical/molecular predictors vs. adjusted AUROC= 0.84 for the SCNP classifier). Finally, when the nodes in the best BBLRS model developed on data from BM samples were used to model read outs from the paired PB samples, the adjusted AUROC of the resulting BBLRS model was comparable to that of the model fit to BM samples. Conclusion: This training set data show the value of performing quantitative SCNP under modulated conditions as the basis for developing highly predictive tests for response to induction chemotherapy. Most importantly, the predictions made by the SCNP classifier are independent of established prognostic factors, such as age and cytogenetics The ability of one set of nodes to accurately predict response in paired BM or PB samples from individual patients suggests that the predictive power of the SCNP assay is independent of sample source, further improving the practicality of the test. Independent validation studies are ongoing. Disclosures: Cesano: Nodality Inc.: Employment, Equity Ownership. Putta:Nodality Inc.: Employment, Equity Ownership. Gayko:Nodality Inc.: Employment, Equity Ownership. Hackett:Nodality Inc.: Consultancy. Rosen:Nodality Inc.: Employment, Equity Ownership.

Abstract 2693 Background: Single Cell Network Profiling (SCNP) is used to measure simultaneously the effects of multiple modulators (including drugs) on signaling cascades at the single cell level. Using this technology, ECOG in collaboration with Nodality is developing several novel biomarker assays with the aim to find blast functional signaling profiles predictive of response to induction therapy and risk of relapse in AML patients. To date, such assays utilized patient bone marrow (BM) as the sample source of blasts. However, in about 65% of patients with AML, circulating peripheral blasts are detected and peripheral blood (PB) sampling is easier and less invasive for patients than BM sampling. Objectives: The objective of this study was to compare by SCNP the functional effects of a panel of compounds simultaneously on different signaling pathways (such as the phosphoinositide 3-kinase (P13K )and the Janus Kinases (Jak) signal transducers and activators of transcription (Stat) pathway) relevant both to the biology of the disease and the development of new therapeutics, in paired, diagnostic, cryopreserved PB mononuclear cells (PBMC) and BMMC samples from 44 AML patients. A paired sample was defined as a BMMC and PBMC specimen collected from the same patient on the same day. Methods: Modulated SCNP using a multiparametric flow cytometry platform was used to evaluate the activation state of intracellular signaling molecules in leukemic blasts under basal conditions and after treatment with specific modulators (Table 1). The SCNP phosphoflow assay was performed on 88 BMMC/PBMC pairs from ECOG trial, E3999. The relationship between readouts of modulated intracellular proteins (“nodes”) between BMMC and PBMC was assessed using linear regression, Bland-Altman method or Lin's concordance correlation coefficient. Results: Table 1 shows the goodness of fit (R2) values from the linear regression analysis for both the basal levels and the modulated levels of intracellular signaling proteins. Most of the signaling nodes show strong correlations (R2 >0.64) with several of the exceptions belonging to nodes with weak response to modulation (e.g. SCF -> p-Akt) or antibodies with dim fluorphores (i.e. Alexa 647). The lack of response is however, consistent between the tissue types for the weak response nodes. Using a rank based metric that is less sensitive to the absolute intensity levels seem to perform better for the antibodies with dim fluorophores. Results from other methods; Bland Altman and Lin's Concordance also show good concordance between the tissue types. Conclusions: The data presented here demonstrate: 1) Specimen source (BM or PB) does not affect proteomic signaling in patients with AML and circulating blasts. 2) PB myeloblasts can be used as a sample source for Nodality SCNP assays to identify functionally distinct leukemic blats cell populations with distinct sensitivities to therapy. 3) The ability to use PB as a sample source will greatly improve the utility of these assays. In particular, our results will facilitate the monitoring of cellular signaling effects following the administration of targeted therapies, e.g., kinase inhibitors, at time-points when BM aspirates are not clinically justifiable. Disclosures: Cesano: Nodality Inc.: Employment, Equity Ownership. Rosen:Nodality Inc.: Employment, Equity Ownership. Putta:Nodality Inc.: Employment, Equity Ownership. Gayko:Nodality Inc.: Employment, Equity Ownership.