Primary myelofibrosis (PMF), together with polycythemia vera (PV) and essential thrombocythemia (ET), belongs to the group of related hematologic cancers named classic Philadelphia-negative myeloproliferative neoplasms (MPNs). PV and ET can evolve to myelofibrosis giving rise to post-PV (PPV-MF) and post-ET (PET-MF) myelofibrosis, which are both defined as secondary myelofibrosis (SMF). Despite the differences, PMF and SMF patients are currently managed in the same way, and risk stratification is based mainly on clinical features and the presence of driver mutations. None of the existing models for MF (e.g. DIPSS, MIPSS70) integrates transcriptomic data. On the other hand, interest has grown in the last few years concerning the ability of gene expression profiles (GEPs) to provide valuable prognostic information. Several studies demonstrated that GEP can improve risk classification in other hematologic malignancies. Therefore, there is a need to better characterize the transcriptomic profile of myelofibrosis in order to add more robustness to the current scoring systems. The main scope of this project was to identify a molecular signature and to build a robust classification model able to distinguish “high risk” MF patients with inferior overall survival from “low risk” ones. We analyzed the gene expression profiles of granulocytes isolated from 114 patients with MF. Cox regression analysis led to the identification of a list of 832 survival-related transcripts characterizing patients who are at high risk for death. Nearest shrunken centroids, subsequent iterations, and k-fold cross-validation were used to build, optimize and validate a classification model, obtaining a final model based on 273 transcripts. Classification of the 114 samples of our dataset with this model resulted in 54 high-risk and 60 low-risk samples. High-risk patients displayed an inferior overall survival and leukemia-free survival. In addition, we observed significant enrichment, within the high-risk group, of clinical and molecular detrimental features included in contemporary prognostic models. Strikingly, several patients belonging to the low and intermediate-1 categories of existing prognostic scores were classified as high-risk with our model. These patients were deceased or leukemia transformed earlier than the prognostic class reference median survival. Moreover, our model showed good performance particularly in distinguishing high-risk and low-risk patients within DIPSS and MIPSS70 intermediate categories. It is noteworthy that intermediate-risk classes represent the most challenging patients’ categories, for whom determining the optimal therapeutic strategy is more difficult. Additionally, to assess if our model was able to improve the prognostic power of current scoring systems, we designed two new combined models by integrating information from our gene expression-based classification within two existing scores (DIPSS and MIPSS70). The Akaike information criterion (AIC) score was used to compare models for prediction of survival. It turned out that both our new combined models showed better AIC values than DIPSS and MIPSS70 alone. Overall, these results demonstrate that GEPs in MF patients correlate with their molecular and clinical features, particularly their survival. Thus suggesting that the evaluation of granulocytes’ gene expression profiles can improve prognostication, particularly with the identification of MF patients’ subgroups characterized by poor prognosis, allowing these patients to be directed towards the most appropriate therapeutic option. These results should be validated in an independent dataset to confirm their predictive power.

La mielofibrosi primaria (PMF), la policitemia vera (PV) e la trombocitemia essenziale (ET), appartengono al gruppo di tumori ematologici correlati denominati neoplasie mieloproliferative classiche Philadelphia negative (MPNs). PV e ET possono evolvere in mielofibrosi (MF), dando origine a MF post-PV (PPV-MF) e post-ET (PET-MF), che sono entrambe definite come MF secondarie (SMF). Nonostante le differenze, i pazienti con PMF e SMF sono gestiti nello stesso modo e la stratificazione del rischio si basa perlopiù su caratteristiche cliniche e mutazioni driver. Nessuno dei modelli esistenti per MF (es. DIPSS e MIPSS70) integra dati trascrittomici. D'altra parte, negli ultimi anni l’interesse verso la capacità dei profili di espressione genica (GEP) di fornire informazioni prognostiche è aumentato. Diversi studi hanno dimostrato che i GEP possono migliorare la classificazione del rischio in altre neoplasie ematologiche. Pertanto, si evidenzia la necessità di migliorare la caratterizzazione trascrittomica della MF al fine di aggiungere robustezza ai sistemi di scoring attualmente usati. Lo scopo principale di questo progetto era quello di identificare una firma molecolare e di costruire un modello di classificazione robusto per distinguere pazienti con MF ad alto rischio, con una sopravvivenza globale inferiore, da pazienti a basso rischio. Abbiamo analizzato i GEP di granulociti isolati da 114 pazienti con MF. La regressione di Cox ha portato all'identificazione di 832 trascritti correlati con la sopravvivenza, caratterizzanti i pazienti ad alto rischio di morte. Nearest shrunken centroids, iterazioni successive e k-fold cross-validation sono stati utilizzati per costruire, ottimizzare e convalidare un modello di classificazione, ottenendo un modello finale basato su 273 trascritti. La classificazione dei 114 campioni ha prodotto 54 campioni ad alto rischio (HR) e 60 campioni a basso rischio (LR). I pazienti HR hanno mostrato sopravvivenza globale e libera da leucemia inferiore. È stato osservato un arricchimento, nel gruppo HR, di caratteristiche cliniche e molecolari dannose incluse nei modelli prognostici attuali. Diversi pazienti low e intermediate-1 secondo gli score prognostici attuali, ma HR secondo il nostro modello, sono deceduti o hanno subito una trasformazione leucemica in un tempo inferiore rispetto alla sopravvivenza mediana di riferimento. Inoltre, il nostro modello ha mostrato buone prestazioni nel distinguere pazienti HR e LR nelle categorie intermedie di DIPSS e MIPSS70. È interessante notare che le classi di rischio intermedio rappresentano le categorie di pazienti più complesse, per le quali è più difficile determinare la strategia terapeutica ottimale. In aggiunta, per valutare se il nostro modello è in grado di migliorare il potere prognostico dei sistemi di scoring attuali, abbiamo progettato due nuovi modelli combinati, integrando le informazioni provenienti dalla nostra classificazione basata su GEP all'interno di due score esistenti (DIPSS e MIPSS70). Per confrontare i modelli è stato usato il criterio d'informazione di Akaike (AIC). È risultato che entrambi i nuovi modelli hanno mostrato valori di AIC migliori rispetto a DIPSS e MIPSS70 da soli. Nel complesso questi risultati dimostrano che i GEP nei pazienti con MF correlano con le loro caratteristiche molecolari e cliniche, in particolare con la loro sopravvivenza. Suggerendo, così, che la valutazione dei GEP dei granulociti può migliorare l'inquadramento prognostico attraverso l’identificazione di sottogruppi di pazienti caratterizzati da prognosi sfavorevole, che possono essere indirizzati verso l’opzione terapeutica più appropriata. Questi risultati dovrebbero essere validati in un dataset indipendente per confermarne il potere predittivo.

Sviluppo di un modello computazionale di analisi di espressione genica per la stratificazione del rischio in pazienti con mielofibrosi / Sara Castellano , 2022 May 27. 34. ciclo, Anno Accademico 2020/2021.

Sviluppo di un modello computazionale di analisi di espressione genica per la stratificazione del rischio in pazienti con mielofibrosi

CASTELLANO, SARA
2022

Abstract

Primary myelofibrosis (PMF), together with polycythemia vera (PV) and essential thrombocythemia (ET), belongs to the group of related hematologic cancers named classic Philadelphia-negative myeloproliferative neoplasms (MPNs). PV and ET can evolve to myelofibrosis giving rise to post-PV (PPV-MF) and post-ET (PET-MF) myelofibrosis, which are both defined as secondary myelofibrosis (SMF). Despite the differences, PMF and SMF patients are currently managed in the same way, and risk stratification is based mainly on clinical features and the presence of driver mutations. None of the existing models for MF (e.g. DIPSS, MIPSS70) integrates transcriptomic data. On the other hand, interest has grown in the last few years concerning the ability of gene expression profiles (GEPs) to provide valuable prognostic information. Several studies demonstrated that GEP can improve risk classification in other hematologic malignancies. Therefore, there is a need to better characterize the transcriptomic profile of myelofibrosis in order to add more robustness to the current scoring systems. The main scope of this project was to identify a molecular signature and to build a robust classification model able to distinguish “high risk” MF patients with inferior overall survival from “low risk” ones. We analyzed the gene expression profiles of granulocytes isolated from 114 patients with MF. Cox regression analysis led to the identification of a list of 832 survival-related transcripts characterizing patients who are at high risk for death. Nearest shrunken centroids, subsequent iterations, and k-fold cross-validation were used to build, optimize and validate a classification model, obtaining a final model based on 273 transcripts. Classification of the 114 samples of our dataset with this model resulted in 54 high-risk and 60 low-risk samples. High-risk patients displayed an inferior overall survival and leukemia-free survival. In addition, we observed significant enrichment, within the high-risk group, of clinical and molecular detrimental features included in contemporary prognostic models. Strikingly, several patients belonging to the low and intermediate-1 categories of existing prognostic scores were classified as high-risk with our model. These patients were deceased or leukemia transformed earlier than the prognostic class reference median survival. Moreover, our model showed good performance particularly in distinguishing high-risk and low-risk patients within DIPSS and MIPSS70 intermediate categories. It is noteworthy that intermediate-risk classes represent the most challenging patients’ categories, for whom determining the optimal therapeutic strategy is more difficult. Additionally, to assess if our model was able to improve the prognostic power of current scoring systems, we designed two new combined models by integrating information from our gene expression-based classification within two existing scores (DIPSS and MIPSS70). The Akaike information criterion (AIC) score was used to compare models for prediction of survival. It turned out that both our new combined models showed better AIC values than DIPSS and MIPSS70 alone. Overall, these results demonstrate that GEPs in MF patients correlate with their molecular and clinical features, particularly their survival. Thus suggesting that the evaluation of granulocytes’ gene expression profiles can improve prognostication, particularly with the identification of MF patients’ subgroups characterized by poor prognosis, allowing these patients to be directed towards the most appropriate therapeutic option. These results should be validated in an independent dataset to confirm their predictive power.
A gene expression-based computational model to improve risk stratification in myelofibrosis
27-mag-2022
TAGLIAFICO, Enrico
File in questo prodotto:
File Dimensione Formato  
Tesi_PhD_Sara_Castellano.pdf

Open access

Descrizione: Tesi definitiva Castellano Sara
Tipologia: Tesi di dottorato
Dimensione 4.2 MB
Formato Adobe PDF
4.2 MB Adobe PDF Visualizza/Apri
Pubblicazioni consigliate

Licenza Creative Commons
I metadati presenti in IRIS UNIMORE sono rilasciati con licenza Creative Commons CC0 1.0 Universal, mentre i file delle pubblicazioni sono rilasciati con licenza Attribuzione 4.0 Internazionale (CC BY 4.0), salvo diversa indicazione.
In caso di violazione di copyright, contattare Supporto Iris

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11380/1278340
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus ND
  • ???jsp.display-item.citation.isi??? ND
social impact