S100 TREATMENT-DEPENDENCE OF HIGH-RISK GENE EXPRESSION SIGNATURES IN DE NOVO FOLLICULAR LYMPHOMA

Abstract A subset of patients (pts.) with follicular lymphoma (FL) will transform to a more aggressive histological sub-type, most typically diffuse large B-cell lymphoma (DLBCL). In general response to therapy is poor and survival short. Paired analysis of samples pre- and post transformation suggest that the molecular mechanisms underlying transformation (Tx) are heterogeneous. In order to independently validate recurring changes in gene/protein expression at transformation (GC phenotype of TxDLBCL, Davies et al., 2002; loss of follicular dendritic Cell (FDC) markers, Shiozawa et al., 2003) a Tx-tissue microarray (Tx-TMA) was created comprising serial samples from 35 pts. (median age 54yrs (22–81) at the time of transformation). In these pts. transformation occurred a median of 3.1years from diagnosis (range 0–15.4) and for each pt. ‘set’ at least 1 pre-Tx FL sample (1–3; n=56), and 1 (1–4; n=44) post transformation sample were represented on the array. To ensure that the Tx-TMA cores accurately represented the corresponding full tissue sections a panel of routine immunohistochemical (IHC) diagnostic markers (n=9) were scored. The concordance between Tx-TMA and full sections (n=10) was >90%. The Tx-TMA was then used to investigate the phenotype of transformed DLBCL, according to the germinal centre (GC)/non-GC like model of de novo DLBCL. Using CD10, BCL6 and MUM1 expression to discriminate between the two subclasses of DLBCL the methodology was first validated on a de novo DLBCL TMA (n=31; 20/31 (65%) non-GC, 11/31 (35%) GC phenotype; 5-yr survival for non-GC pts. 51% and for GC pts., 73%). IHC confirmed the results of gene expression profiling indicating that in 31/35 (89%) pts. transformed DLBCL was of GC phenotype (28/35 (80%) CD10+ and 3/35 (9%) CD10-, BCL6+, MUM1-). Of the remainder, 4/35 were CD10-, BCL6+, of which 3/4 were MUM1+ (3/35 (9%) non-GC phenotype; 1/4 MUM1 was not assessable). Similarly the Tx-TMA confirmed loss of FDC markers (CD21 and CD23) on transformation. Samples from 28 pts were evaluable for CD21 and CD23 IHC expression. In 71% (20/28) of pts. the FDC meshwork was lost or became more sparse on transformation (CD21 loss 15/28 (54%); CD23 loss 17/28 (61%)). The most discriminating changes in gene expression on transformation are now being assessed by IHC. Aurora kinase B (ARKB) is an attractive therapeutic target given that disruption of ARK function results in the induction of apoptosis in RL, a t(14:18) positive DLBCL cell line (Harrington et al. 2004). The observed elevation in ARKB transcription on transformation was confirmed by IHC in this series. The Tx-TMA showed ARKB expression increased on Tx in 13/33 (40%) pts., potentially defining a subset of pts. who might be considered for ARKB directed therapies. Expression of ARKB was low throughout in 18/33 (55%) pts and decreased in 2/33 (6%) pts.; difference in ARKB expression was not significantly associated with survival. These preliminary studies suggest that the availability of TMA of serial biopsies from pts. with transformed FL will provide a powerful means of assessing the relevance of gene expression, both within the tumour and the microenvironment while facilitating the selection of patients most likely to benefit from directed therapeutic approaches.

Download Full-text

Microarray Classification of Myelodysplastic Syndrome (MDS) Identifies Subgroups with Distinct Clinical Outcomes.

Blood ◽

10.1182/blood.v110.11.2426.2426 ◽

2007 ◽

Vol 110 (11) ◽

pp. 2426-2426

Author(s):

Ken I. Mills ◽

Alex Kohlmann ◽

Mickey Williams ◽

Wei-Min Liu ◽

Rachel Li ◽

...

Keyword(s):

Gene Expression ◽

De Novo ◽

Cell Signalling ◽

Bioinformatic Analysis ◽

Initial Study ◽

Outcome Data ◽

Kaplan Meier ◽

Significant Difference ◽

Gene Expression Signatures ◽

Microarray Classification

Abstract The MILE (Microarray Innovations in LEukemia) study has previously shown that gene expression signatures associated with initial leukaemia classifier (LCver7) give an overall cross-validation accuracy of >95% for distinct sub-classes of pediatric and adult leukemias. However, only 50% of the 174 MDS samples in the whole-genome microarray analysis (Stage 1) of the MILE study were correctly identified; the remainder showed AML-like or non-leukemia-like gene profiles. An external morphological review (DB & HL) according to FAB and WHO criteria, of the 174 slides was performed independently (blind) which resulted in 6 samples being reclassified as AML and 4 non-leukemia cases excluded from the study. A recently improved, hierarchical based algorithm correctly identified 100% of the confirmed MDS cases. In this study, using LCver7, the confirmed 164 samples had 50% MDS classifications (Class 17), 23.8% non-leukemia classifications (Class 18), and 22.6% AML classifications (Classes 13 or 14) with the remaining 3.7% having a classification tie between 2 or 3 Classes (due to low confidence). No 5q- syndrome patients had an AML call, whilst 68.3% of RAEB2 patients had an AML classification and none were Class 18. Similarly, 95.6% of Low IPSS patients were classified as Class 17 or 18, whilst all patients (n=5) with High IPSS had an AML call. The classification was independent of blast cells: 10.2% of Class 18 calls had >5% blasts; 28.2% of AML-like cases had <5% blasts. Outcome data (132 MDS patients) was correlated with Class: significant difference (p<0.028, (Kaplan-Meier)) was seen in overall survival; with p <0.004 if AML (Classes 13 & 14) was compared to “non AML” (Classes 17 & 18). Statistically significant differences were seen for time to transformation to AML between the classes (p<0.0001) and between AML and “non AML” (p<0.00007, Kaplan-Meier)) with a probability of transformation of 44% at 18 months for the AML group compared to <8% for the “non-AML” group. A further linear classifier has been used to discriminate patients who transform to AML within 18 months (poor prognosis) with patients with no transformation after >60 months (good prognostic group). Bioinformatic analysis of molecular mapped functions and canonical pathways showed that cell signalling processes were over-represented when comparing de novo AML (n=204) with MDS, from the MILE study, whilst signal transduction pathways were deregulated when comparing non-leukemia samples (n=71) with MDS. Similar pathways and functions were also deregulated when comparing the correctly classified MDS with Class 17 call against MDS with Class 18 call and MDS with AML Classes 13 or 14 calls. In conclusion, the use of microarrays within the initial study, solely intended for diagnostic purposes, has now evolved towards a position in which novel prognostic value may be gained from distinct gene expression signatures. This has also resulted in a better molecular understanding of the progression from non-leukemia, through MDS into full blown AML.

Download Full-text

Systematic Drug Repositioning By Integrating Transcriptome and Historical Clinical Data, Identification of Digoxin As a Novel Drug Reposition Candidate for High-Risk Myelodysplastic Syndromes

Blood ◽

10.1182/blood.v126.23.4118.4118 ◽

2015 ◽

Vol 126 (23) ◽

pp. 4118-4118 ◽

Cited By ~ 1

Author(s):

Huojun Cao ◽

Zheng Yin ◽

Shenyi Chen ◽

Timothy Liu ◽

Jiyong Liu ◽

...

Keyword(s):

Gene Expression ◽

High Risk ◽

Data Warehouse ◽

Clinical Data ◽

Drug Repositioning ◽

Second Stage ◽

Clinical Data Warehouse ◽

Gene Expression Signatures ◽

Drug Reposition ◽

Meteor Data

Abstract Background: Traditional drug development typically takes more than 15 years and costs 1 billions to bring a new drug to market. Drug repositioning, the application of established drug compounds to new therapeutic indications can hasten effective therapies at a lower cost. Although a number of treatment options are available, there is currently no proven cure for Myelodysplastic syndromes (MDS) besides bone marrow transplantation. We employed a novel two stages, transdisciplinary method to identify potential drug reposition candidates for MDS with commonly prescribed drugs. We found Digoxin, a drug prescribed for heart failure, as the leading candidate. Methods: After we downloaded all human mRNA microarray platforms and their associated annotation files from NCBI GEO database (as Jul 2014), we retained 145 microarray platforms from more than 4k unique Gene IDs. We then used a text mining tool, MetaMap (http://metamap.nlm.nih.gov/), to map text to Unified Medical Language System (UMLS) Concept Unique Identifier (CUI). Only concepts belong to Anatomical Abnormality, Disease or Syndrome and Neoplastic Process were kept. Then we manually grouped microarray samples into healthy and pathological conditions and susing R package, Limma, we generated diseases signatures by systematically comparing with 1.3M drug response gene expression signatures of 20K drugs/compounds and using pattern matching algorithm. In the second stage, drug repositioning candidates for MDS were further analyzed by integrating historical clinical data warehouse, METEOR (M ethodist E nvironment for T ranslational E nhancement and O utcomes R esearch). METEOR data warehouse contains clinical records of Houston Methodist Hospital System and external data dating from Jan 1st 2006 to present (2015). There are over 1 million unique patients and over 4 million unique patient encounters in METEOR. Selected drug repositioning candidates predicted by transcriptome data in the first stage were tested for their effects on MDS patients using METEOR data. Age, gender, race matched myelodysplastic syndrome (MDS) patients that have taken or not taken a specific drug were compared for their overall survival. Results: 1174 diseases signatures for 301 human diseases were generated; for each the signatures, we computed it's repositioning score in CMap and LINCS database by non-parametric rank-ordered kolmogorov-Smirnov statistics. The overall disease-drug reposition map is presented in Figure 1. Many FDA approved indications were recovered by our disease-drug reposition map including 5'-azacitidine which was at top 1% of leading hints in the disease-drug reposition map (Figure 1B). We also find some potential new indication for old drugs such as cardiac glycosides (Digoxin, Digitoxin, Lanatoside C and others) for MDS (Figure 1D). In the second stage we found Digoxin, had a significant positive effect on high risk patients (Figure 2) unlike those for low risk MDS. Conclusions: We have generated 1174 gene expression signatures for 301 human diseases including MDS and systematically compared them with gene expression signatures of 20K drug/compounds. By integrating clinical data warehouse, METEOR, we further verified one potential drug repositioning candidate, Digoxin for high-risk MDS. We are currently testing Digoxin like compounds for its effects in short-term bone marrow culture from high-risk MDS patients. Preliminary results show encouraging results which will be reported at the Annual ASH meeting 2015. Disclosures No relevant conflicts of interest to declare.

Download Full-text

Prognostic relevance of CD4+ T-cells in the microenvironment of newly diagnosed follicular lymphoma (FL) patients is independent of the tumor gene expression profile.

Journal of Clinical Oncology ◽

10.1200/jco.2020.38.15_suppl.8052 ◽

2020 ◽

Vol 38 (15_suppl) ◽

pp. 8052-8052

Author(s):

Patrizia Mondello ◽

Angelo Fama ◽

Melissa C. Larson ◽

Andew L. Feldman ◽

Zhi-Zhang J Yang ◽

...

Keyword(s):

Gene Expression ◽

T Cell ◽

High Risk ◽

Follicular Lymphoma ◽

Prognostic Score ◽

Cd4 T Cell ◽

Newly Diagnosed ◽

Early Failure ◽

Tumor Gene Expression ◽

Tumor Gene

8052 Background: A significant proportion of patients with FL experience an early relapse and a subsequent poor outcome. While several prognostic indices have been developed, none were designed to predict early failure. Recently, we established that lack of intrafollicular CD4+ T-cell expression predicted risk of early failure, and integrating this microenvironment biomarker with the Follicular Lymphoma International Prognostic Index, termed BioFLIPI, further improved identification of FL patients at risk of early failure ( Blood 2019;134(suppl1):121). However, the microenvironment may be influenced by the genetic composition of tumor. We investigated whether the CD4 biomarker and BioFLIPI were impacted by genetic features of the tumor as assessed by a 23-gene expression prognostic score ( Lancet Oncol 2018;19:549-61). Methods: Of the 186 cases with FL grade 1-3A treated with immunochemotherapy (IC) in our prior study, 152 had digital expression quantification of 23 selected genes (23-GEP score), which used RNA from formalin-fixed, paraffin-embedded samples. Event-free survival (EFS) was defined as time from diagnosis to progression, relapse, retreatment, or death. Early failure was defined as failing to achieve EFS at 24 months. Risk of early failure was estimated using odds ratios (ORs) and 95% confidence intervals from logistic regression models. We also used Cox regression to assess associations with continuous EFS and overall survival (OS). Results: 28% of patients failed to achieve EFS24. Lack of CD4+ intrafollicular expression (38% of patients, OR = 2.33, p = 0.024) and high risk 23-GEP score (26% of patients, OR = 3.52, p = 0.001) each predicted early failure, and in a multivariable model that included FLIPI, both CD4+ (OR = 2.26, p = 0.046) and 23-GEP score (OR = 2.26, p = 0.0.057) remained predictors. Similarly, BioFLIPI modeled as a continuous score (1-4, OR per one point increase = 2.31, p < 0.001) predicted early failure, and the association remained (OR = 2.14, p < 0.001) when the high risk 23-GEP score (OR = 2.79, p = 0.013) was included in the model. When stratified on 23-GEP score, BioFLIPI was a stronger predictor of early failure in low risk (74%, OR = 2.51, p = 0.002) relative to high risk (26%, OR = 1.55, p = 0.27) patients. Similar patterns were observed for EFS and OS. Conclusions: CD4+ T-cell infiltrate and tumor gene expression appear to be independently predictive of early failure in newly diagnosed FL patients treated with IC. Future studies should integrate and validate these measures.

Download Full-text

De Novo Generation of Hit-like Molecules from Gene Expression Signatures Using Artificial Intelligence

10.26434/chemrxiv.7294388.v1 ◽

2018 ◽

Cited By ~ 4

Author(s):

Oscar Mendez-Lucio ◽

Benoit Baillif ◽

Djork-Arné Clevert ◽

David Rouquié ◽

Joerg Wichard

Keyword(s):

Gene Expression ◽

Artificial Intelligence ◽

De Novo ◽

Molecular Design ◽

Gene Expression Signature ◽

Cellular Level ◽

Generative Models ◽

Generative Adversarial Network ◽

Adversarial Network ◽

Gene Expression Signatures

Finding new molecules with a desired biological activity is an extremely difficult task. In this context, artificial intelligence and generative models have been used for molecular <i>de novo</i> design and compound optimization. Herein, we report the first generative model that bridges systems biology and molecular design conditioning a generative adversarial network with transcriptomic data. By doing this we could generate molecules that have high probability to produce a desired biological effect at cellular level. We show that this model is able to design active-like molecules for desired targets without any previous target annotation of the training compounds as long as the gene expression signature of the desired state is provided. The molecules generated by this model are more similar to active compounds than the ones identified by similarity of gene expression signatures, which is the state-of-the-art method for navigating compound-induced gene expression data. Overall, this method represents a novel way to bridge chemistry and biology to advance in the long and difficult road of drug discovery.

Download Full-text