Large registry-based analysis of genetic predisposition to tuberculosis identifies genetic risk factors at HLA.

Tuberculosis is a significant public health concern resulting in the death of over 1 million individuals each year worldwide. While treatment options and vaccines exist, a substantial number of infections still remain untreated or are caused by treatment resistant strains. Therefore, it is important to identify mechanisms that contribute to risk and prognosis of tuberculosis as this may provide tools to understand disease mechanisms and provide novel treatment options for those with severe infection. Our goal was to identify genetic risk factors that contribute to the risk of tuberculosis and to understand biological mechanisms and causality behind the risk of tuberculosis. A total of 1895 individuals in the FinnGen study had International Classification of Diseases-based tuberculosis diagnosis. Genome-wide association study analysis identified genetic variants with statistically significant association with tuberculosis at the human leukocyte antigen (HLA) region (P < 5e-8). Fine mapping of the HLA association provided evidence for one protective haplotype tagged by HLA DQB1*05:01 (P = 1.82E-06, OR = 0.81 [CI 95% 0.74-0.88]), and predisposing alleles tagged by HLA DRB1*13:02 (P = 0.00011, OR = 1.35 [CI 95% 1.16-1.57]). Furthermore, genetic correlation analysis showed association with earlier reported risk factors including smoking (P < 0.05). Mendelian randomization supported smoking as a risk factor for tuberculosis (inverse-variance weighted P < 0.05, OR = 1.83 [CI 95% 1.15-2.93]) with no significant evidence of pleiotropy. Our findings indicate that specific HLA alleles associate with the risk of tuberculosis. In addition, lifestyle risk factors such as smoking contribute to the risk of developing tuberculosis.

Erythroid/megakaryocytic differentiation confers BCL-XL dependency and venetoclax resistance in acute myeloid leukemia.

Myeloid neoplasms with erythroid or megakaryocytic differentiation include pure erythroid leukemia, myelodysplastic syndrome with erythroid features, and acute megakaryoblastic leukemia (FAB M7) and are characterized by poor prognosis and limited treatment options. Here, we investigate the drug sensitivity landscape of these rare malignancies. We show that acute myeloid leukemia (AML) cells with erythroid or megakaryocytic differentiation depend on the antiapoptotic protein B-cell lymphoma (BCL)-XL, rather than BCL-2, using combined ex vivo drug sensitivity testing, genetic perturbation, and transcriptomic profiling. High-throughput screening of >500 compounds identified the BCL-XL-selective inhibitor A-1331852 and navitoclax as highly effective against erythroid/megakaryoblastic leukemia cell lines. In contrast, these AML subtypes were resistant to the BCL-2 inhibitor venetoclax, which is used clinically in the treatment of AML. Consistently, genome-scale CRISPR-Cas9 and RNAi screening data demonstrated the striking essentiality of BCL-XL-encoding BCL2L1 but not BCL2 or MCL1, for the survival of erythroid/megakaryoblastic leukemia cell lines. Single-cell and bulk transcriptomics of patient samples with erythroid and megakaryoblastic leukemias identified high BCL2L1 expression compared with other subtypes of AML and other hematological malignancies, where BCL2 and MCL1 were more prominent. BCL-XL inhibition effectively killed blasts in samples from patients with AML with erythroid or megakaryocytic differentiation ex vivo and reduced tumor burden in a mouse erythroleukemia xenograft model. Combining the BCL-XL inhibitor with the JAK inhibitor ruxolitinib showed synergistic and durable responses in cell lines. Our results suggest targeting BCL-XL as a potential therapy option in erythroid/megakaryoblastic leukemias and highlight an AML subgroup with potentially reduced sensitivity to venetoclax-based treatments.

Therapeutic targeting of LCK tyrosine kinase and mTOR signaling in T-cell acute lymphoblastic leukemia.

Relapse and refractory T-cell acute lymphoblastic leukemia (T-ALL) has a poor prognosis, and new combination therapies are sorely needed. Here, we used an ex vivo high-throughput screening platform to identify drug combinations that kill zebrafish T-ALL and then validated top drug combinations for preclinical efficacy in human disease. This work uncovered potent drug synergies between AKT/mTORC1 (mammalian target of rapamycin complex 1) inhibitors and the general tyrosine kinase inhibitor dasatinib. Importantly, these same drug combinations effectively killed a subset of relapse and dexamethasone-resistant zebrafish T-ALL. Clinical trials are currently underway using the combination of mTORC1 inhibitor temsirolimus and dasatinib in other pediatric cancer indications, leading us to prioritize this therapy for preclinical testing. This combination effectively curbed T-ALL growth in human cell lines and primary human T-ALL and was well tolerated and effective in suppressing leukemia growth in patient-derived xenografts (PDX) grown in mice. Mechanistically, dasatinib inhibited phosphorylation and activation of the lymphocyte-specific protein tyrosine kinase (LCK) to blunt the T-cell receptor (TCR) signaling pathway, and when complexed with mTORC1 inhibition, induced potent T-ALL cell killing through reducing MCL-1 protein expression. In total, our work uncovered unexpected roles for the LCK kinase and its regulation of downstream TCR signaling in suppressing apoptosis and driving continued leukemia growth. Analysis of a wide array of primary human T-ALLs and PDXs grown in mice suggest that combination of temsirolimus and dasatinib treatment will be efficacious for a large fraction of human T-ALLs.

Ex vivo venetoclax sensitivity testing predicts treatment response in acute myeloid leukemia.

The BCL-2 inhibitor venetoclax has revolutionized the treatment of acute myeloid leukemia (AML) in patients not benefiting from intensive chemotherapy. Nevertheless, treatment failure remains a challenge, and predictive markers are needed, particularly for relapsed or refractory AML. Ex vivo drug sensitivity testing may correlate with outcomes, but its prospective predictive value remains unexplored. Here we report the results of the first stage of the prospective phase II VenEx trial evaluating the utility and predictiveness of venetoclax sensitivity testing using different cell culture conditions and cell viability assays in patients receiving venetoclax-azacitidine. Participants with de novo AML ineligible for intensive chemotherapy, relapsed or refractory AML, or secondary AML were included. The primary endpoint was the treatment response in participants showing ex vivo sensitivity and the key secondary endpoints were the correlation of sensitivity with responses and survival. Venetoclax sensitivity testing was successful in 38/39 participants. Experimental conditions significantly influenced the predictive accuracy. Blast-specific venetoclax sensitivity measured in conditioned medium most accurately correlated with treatment outcomes; 88% of sensitive participants achieved a treatment response. The median survival was significantly longer for participants who were ex vivo-sensitive to venetoclax (14.6 months for venetoclax-sensitive patients vs. 3.5 for venetoclax-insensitive patients, P<0.001). This analysis illustrates the feasibility of integrating drug-response profiling into clinical practice and demonstrates excellent predictivity. This trial is registered with ClinicalTrials.gov identifier: NCT04267081.

Using Proteomics Data to Identify Personalized Treatments in Multiple Myeloma: A Machine Learning Approach.

This paper describes a machine learning (ML) decision support system to provide a list of chemotherapeutics that individual multiple myeloma (MM) patients are sensitive/resistant to, based on their proteomic profile. The methodology used in this study involved understanding the parameter space and selecting the dominant features (proteomics data), identifying patterns of proteomic profiles and their association to the recommended treatments, and defining the decision support system of personalized treatment as a classification problem. During the data analysis, we compared several ML algorithms, such as linear regression, Random Forest, and support vector machines, to classify patients as sensitive/resistant to therapeutics. A further analysis examined data-balancing techniques that emerged due to the small cohort size. The results suggest that utilizing proteomics data is a promising approach for identifying effective treatment options for patients with MM (reaching on average an accuracy of 81%). Although this pilot study was limited by the small patient cohort (39 patients), which restricted the training and validation of the explored ML solutions to identify complex associations between proteins, it holds great promise for developing personalized anti-MM treatments using ML approaches.

Fusion gene detection by RNA-sequencing complements diagnostics of acute myeloid leukemia and identifies recurring NRIP1-MIR99AHG rearrangements.

Identification of fusion genes in clinical routine is mostly based on cytogenetics and targeted molecular genetics, such as metaphase karyotyping, fluorescence in situ hybridization and reverse-transcriptase polymerase chain reaction. However, sequencing technologies are becoming more important in clinical routine as processing time and costs per sample decrease. To evaluate the performance of fusion gene detection by RNAsequencing compared to standard diagnostic techniques, we analyzed 806 RNA-sequencing samples from patients with acute myeloid leukemia using two state-of-the-art software tools, namely Arriba and FusionCatcher. RNA-sequencing detected 90% of fusion events that were reported by routine with high evidence, while samples in which RNA-sequencing failed to detect fusion genes had overall lower and inhomogeneous sequence coverage. Based on properties of known and unknown fusion events, we developed a workflow with integrated filtering strategies for the identification of robust fusion gene candidates by RNA-sequencing. Thereby, we detected known recurrent fusion events in 26 cases that were not reported by routine and found discrepancies in evidence for known fusion events between routine and RNA-sequencing in three cases. Moreover, we identified 157 fusion genes as novel robust candidates and comparison to entries from ChimerDB or Mitelman Database showed novel recurrence of fusion genes in 14 cases. Finally, we detected the novel recurrent fusion gene NRIP1- MIR99AHG resulting from inv(21)(q11.2;q21.1) in nine patients (1.1%) and LTN1-MX1 resulting from inv(21)(q21.3;q22.3) in two patients (0.25%). We demonstrated that NRIP1-MIR99AHG results in overexpression of the 3' region of MIR99AHG and the disruption of the tricistronic miRNA cluster miR-99a/let-7c/miR-125b-2. Interestingly, upregulation of MIR99AHG and deregulation of the miRNA cluster, residing in the MIR99AHG locus, are known mechanisms of leukemogenesis in acute megakaryoblastic leukemia. Our findings demonstrate that RNA-sequencing has a strong potential to improve the systematic detection of fusion genes in clinical applications and provides a valuable tool for fusion discovery.

Prognostic significance of esterase gene expression in multiple myeloma.

BACKGROUND: Esterase enzymes differ in substrate specificity and biological function and may display dysregulated expression in cancer. This study evaluated the biological significance of esterase expression in multiple myeloma (MM). METHODS: For gene expression profiling and evaluation of genomic variants in the Institute for Molecular Medicine Finland (FIMM) cohort, bone marrow aspirates were obtained from patients with newly diagnosed MM (NDMM) or relapsed/refractory MM (RRMM). CD138+ plasma cells were enriched and used for RNA sequencing and analysis, and to evaluate genomic variation. The Multiple Myeloma Research Foundation (MMRF) Relating Clinical Outcomes in MM to Personal Assessment of Genetic Profile (CoMMpass) dataset was used for validation of the findings from FIMM. RESULTS: MM patients (NDMM, n = 56; RRMM, n = 78) provided 171 bone marrow aspirates (NDMM, n = 56; RRMM, n = 115). Specific esterases exhibited relatively high or low expression in MM, and expression of specific esterases (UCHL5, SIAE, ESD, PAFAH1B3, PNPLA4 and PON1) was significantly altered on progression from NDMM to RRMM. High expression of OVCA2, PAFAH1B3, SIAE and USP4, and low expression of PCED1B, were identified as poor prognostic markers (P < 0.05). The MMRF CoMMpass dataset provided validation that higher expression of PAFAH1B3 and SIAE, and lower expression of PCED1B, were associated with poor prognosis. CONCLUSIONS: Esterase gene expression levels change as patients progress from NDMM to RRMM. High expression of OVCA2, PAFAH1B3, USP4 and SIAE, and low expression of PCED1B, are poor prognostic markers in MM, suggesting a role for these esterases in myeloma biology.

Endogenous and combination retinoids are active in myelomonocytic leukemias.

Retinoid therapy transformed response and survival outcomes in acute promyelocytic leukemia (APL), but has demonstrated only modest activity in non-APL forms of acute myeloid leukemia (AML). The presence of natural retinoids in vivo could influence the efficacy of pharmacologic agonists and antagonists. We found that natural RXRA ligands, but not RARA ligands, were present in murine MLL-AF9-derived myelomonocytic leukemias in vivo and that the concurrent presence of receptors and ligands acted as tumor suppressors. Pharmacologic retinoid responses could be optimized by concurrent targeting RXR ligands (e.g. bexarotene) and RARA ligands (e.g. all-trans retinoic acid, ATRA), which induced either leukemic maturation or apoptosis depending on cell culture conditions. Co-repressor release from the RARA:RXRA heterodimer occurred with RARA activation, but not RXRA activation, providing an explanation for the combination synergy. Combination synergy could be replicated in additional, but not all, AML cell lines and primary samples, and was associated with improved survival in vivo, although tolerability of bexarotene administration in mice remained an issue. These data provide insight into the basal presence of natural retinoids in leukemias in vivo and a potential strategy for clinical retinoid combination regimens in leukemias beyond acute promyelocytic leukemia.

Multi-parametric single cell evaluation defines distinct drug responses in healthy hematologic cells that are retained in corresponding malignant cell types.

Innate drug sensitivity in healthy cells aids identification of lineage specific anti-cancer therapies and reveals off-target effects. To characterize the diversity in drug responses in the major hematopoietic cell types, we simultaneously assessed their sensitivity to 71 small molecules utilizing a multi-parametric flow cytometry assay and mapped their proteomic and basal signaling profiles. Unsupervised hierarchical clustering identified distinct drug responses in healthy cell subsets based on their cellular lineage. Compared to other cell types, CD19+/B and CD56+/NK cells were more sensitive to dexamethasone, venetoclax and midostaurin, while monocytes were more sensitive to trametinib. Venetoclax exhibited dose-dependent cell selectivity that inversely correlated to STAT3 phosphorylation. Lineage specific effect of midostaurin was similarly detected in CD19+/B cells from healthy, acute myeloid leukemia and chronic lymphocytic leukemia samples. Comparison of drug responses in healthy and neoplastic cells showed that healthy cell responses are predictive of the corresponding malignant cell response. Taken together, understanding drug sensitivity in the healthy cell-of-origin provides opportunities to obtain a new level of therapy precision and avoid off-target toxicity.

Drug combination sensitivity scoring facilitates the discovery of synergistic and efficacious drug combinations in cancer.

High-throughput drug screening has facilitated the discovery of drug combinations in cancer. Many existing studies adopted a full matrix design, aiming for the characterization of drug pair effects for cancer cells. However, the full matrix design may be suboptimal as it requires a drug pair to be combined at multiple concentrations in a full factorial manner. Furthermore, many of the computational tools assess only the synergy but not the sensitivity of drug combinations, which might lead to false positive discoveries. We proposed a novel cross design to enable a more cost-effective and simultaneous testing of drug combination sensitivity and synergy. We developed a drug combination sensitivity score (CSS) to determine the sensitivity of a drug pair, and showed that the CSS is highly reproducible between the replicates and thus supported its usage as a robust metric. We further showed that CSS can be predicted using machine learning approaches which determined the top pharmaco-features to cluster cancer cell lines based on their drug combination sensitivity profiles. To assess the degree of drug interactions using the cross design, we developed an S synergy score based on the difference between the drug combination and the single drug dose-response curves. We showed that the S score is able to detect true synergistic and antagonistic drug combinations at an accuracy level comparable to that using the full matrix design. Taken together, we showed that the cross design coupled with the CSS sensitivity and S synergy scoring methods may provide a robust and accurate characterization of both drug combination sensitivity and synergy levels, with minimal experimental materials required. Our experimental-computational approach could be utilized as an efficient pipeline for improving the discovery rate in high-throughput drug combination screening, particularly for primary patient samples which are difficult to obtain.

Improving genomics-based predictions for precision medicine through active elicitation of expert knowledge.

Motivation: Precision medicine requires the ability to predict the efficacies of different treatments for a given individual using high-dimensional genomic measurements. However, identifying predictive features remains a challenge when the sample size is small. Incorporating expert knowledge offers a promising approach to improve predictions, but collecting such knowledge is laborious if the number of candidate features is very large. Results: We introduce a probabilistic framework to incorporate expert feedback about the impact of genomic measurements on the outcome of interest and present a novel approach to collect the feedback efficiently, based on Bayesian experimental design. The new approach outperformed other recent alternatives in two medical applications: prediction of metabolic traits and prediction of sensitivity of cancer cells to different drugs, both using genomic features as predictors. Furthermore, the intelligent approach to collect feedback reduced the workload of the expert to approximately 11%, compared to a baseline approach. Availability and implementation: Source code implementing the introduced computational methods is freely available at https://github.com/AaltoPML/knowledge-elicitation-for-precision-medicine. Supplementary information: Supplementary data are available at Bioinformatics online.

JAK1/2 and BCL2 inhibitors synergize to counteract bone marrow stromal cell-induced protection of AML.

The bone marrow (BM) provides a protective microenvironment to support the survival of leukemic cells and influence their response to therapeutic agents. In acute myeloid leukemia (AML), the high rate of relapse may in part be a result of the inability of current treatment to effectively overcome the protective influence of the BM niche. To better understand the effect of the BM microenvironment on drug responses in AML, we conducted a comprehensive evaluation of 304 inhibitors, including approved and investigational agents, comparing ex vivo responses of primary AML cells in BM stroma-derived and standard culture conditions. In the stroma-based conditions, the AML patient cells exhibited significantly reduced sensitivity to 12% of the tested compounds, including topoisomerase II, B-cell chronic lymphocytic leukemia/lymphoma 2 (BCL2), and many tyrosine kinase inhibitors (TKIs). The loss of TKI sensitivity was most pronounced in patient samples harboring FLT3 or PDGFRB alterations. In contrast, the stroma-derived conditions enhanced sensitivity to Janus kinase (JAK) inhibitors. Increased cell viability and resistance to specific drug classes in the BM stroma-derived conditions was a result of activation of alternative signaling pathways mediated by factors secreted by BM stromal cells and involved a switch from BCL2 to BCLXL-dependent cell survival. Moreover, the JAK1/2 inhibitor ruxolitinib restored sensitivity to the BCL2 inhibitor venetoclax in AML patient cells ex vivo in different model systems and in vivo in an AML xenograft mouse model. These findings highlight the potential of JAK inhibitors to counteract stroma-induced resistance to BCL2 inhibitors in AML.

Somatic MED12 Nonsense Mutation Escapes mRNA Decay and Reveals a Motif Required for Nuclear Entry.

MED12 is a key component of the transcription-regulating Mediator complex. Specific missense and in-frame insertion/deletion mutations in exons 1 and 2 have been identified in uterine leiomyomas, breast tumors, and chronic lymphocytic leukemia. Here, we characterize the first MED12 5' end nonsense mutation (c.97G>T, p.E33X) identified in acute lymphoblastic leukemia and show that it escapes nonsense-mediated mRNA decay (NMD) by using an alternative translation initiation site. The resulting N-terminally truncated protein is unable to enter the nucleus due to the lack of identified nuclear localization signal (NLS). The absence of NLS prevents the mutant MED12 protein to be recognized by importin-α and subsequent loading into the nuclear pore complex. Due to this mislocalization, all interactions between the MED12 mutant and other Mediator components are lost. Our findings provide new mechanistic insights into the MED12 functions and indicate that somatic nonsense mutations in early exons may avoid NMD.

The impact of RNA sequence library construction protocols on transcriptomic profiling of leukemia.

BACKGROUND: RNA sequencing (RNA-seq) has become an indispensable tool to identify disease associated transcriptional profiles and determine the molecular underpinnings of diseases. However, the broad adaptation of the methodology into the clinic is still hampered by inconsistent results from different RNA-seq protocols and involves further evaluation of its analytical reliability using patient samples. Here, we applied two commonly used RNA-seq library preparation protocols to samples from acute leukemia patients to understand how poly-A-tailed mRNA selection (PA) and ribo-depletion (RD) based RNA-seq library preparation protocols affect gene fusion detection, variant calling, and gene expression profiling. RESULTS: Overall, the protocols produced similar results with consistent outcomes. Nevertheless, the PA protocol was more efficient in quantifying expression of leukemia marker genes and showed better performance in the expression-based classification of leukemia. Independent qRT-PCR experiments verified that the PA protocol better represented total RNA compared to the RD protocol. In contrast, the RD protocol detected a higher number of non-coding RNA features and had better alignment efficiency. The RD protocol also recovered more known fusion-gene events, although variability was seen in fusion gene predictions. CONCLUSION: The overall findings provide a framework for the use of RNA-seq in a precision medicine setting with limited number of samples and suggest that selection of the library preparation protocol should be based on the objectives of the analysis.

Somatic STAT3 mutations in large granular lymphocytic leukemia.

BACKGROUND: T-cell large granular lymphocytic leukemia is a rare lymphoproliferative disorder characterized by the expansion of clonal CD3+CD8+ cytotoxic T lymphocytes (CTLs) and often associated with autoimmune disorders and immune-mediated cytopenias. METHODS: We used next-generation exome sequencing to identify somatic mutations in CTLs from an index patient with large granular lymphocytic leukemia. Targeted resequencing was performed in a well-characterized cohort of 76 patients with this disorder, characterized by clonal T-cell-receptor rearrangements and increased numbers of large granular lymphocytes. RESULTS: Mutations in the signal transducer and activator of transcription 3 gene (STAT3) were found in 31 of 77 patients (40%) with large granular lymphocytic leukemia. Among these 31 patients, recurrent mutational hot spots included Y640F in 13 (17%), D661V in 7 (9%), D661Y in 7 (9%), and N647I in 3 (4%). All mutations were located in exon 21, encoding the Src homology 2 (SH2) domain, which mediates the dimerization and activation of STAT protein. The amino acid changes resulted in a more hydrophobic protein surface and were associated with phosphorylation of STAT3 and its localization in the nucleus. In vitro functional studies showed that the Y640F and D661V mutations increased the transcriptional activity of STAT3. In the affected patients, downstream target genes of the STAT3 pathway (IFNGR2, BCL2L1, and JAK2) were up-regulated. Patients with STAT3 mutations presented more often with neutropenia and rheumatoid arthritis than did patients without these mutations. CONCLUSIONS: The SH2 dimerization and activation domain of STAT3 is frequently mutated in patients with large granular lymphocytic leukemia; these findings suggest that aberrant STAT3 signaling underlies the pathogenesis of this disease. (Funded by the Academy of Finland and others.).

Discovery of somatic STAT5b mutations in large granular lymphocytic leukemia.

Large granular lymphocytic (LGL) leukemia is characterized by clonal expansion of cytotoxic T cells or natural killer cells. Recently, somatic mutations in the signal transducer and activator of transcription 3 (STAT3) gene were discovered in 28% to 40% of LGL leukemia patients. By exome and transcriptome sequencing of 2 STAT3 mutation-negative LGL leukemia patients, we identified a recurrent, somatic missense mutation (Y665F) in the Src-like homology 2 domain of the STAT5b gene. Targeted amplicon sequencing of 211 LGL leukemia patients revealed 2 additional patients with STAT5b mutations (N642H), resulting in a total frequency of 2% (4 of 211) of STAT5b mutations across all patients. The Y665F and N642H mutant constructs increased the transcriptional activity of STAT5 and tyrosine (Y694) phosphorylation, which was also observed in patient samples. The clinical course of the disease in patients with the N642H mutation was aggressive and fatal, clearly different from typical LGL leukemia with a relatively favorable outcome. This is the first time somatic STAT5 mutations are discovered in human cancer and further emphasizes the role of STAT family genes in the pathogenesis of LGL leukemia.

Statistical detection of quantitative protein biomarkers provides insights into signaling networks deregulated in acute myeloid leukemia.

The increasing coverage and sensitivity of LC-MS/MS-based proteomics have expanded its applications in systems medicine. In particular, label-free quantitation approaches are enabling biomarker discovery in terms of statistical comparison of proteomic profiles across large numbers of clinical samples. However, it still remains poorly understood how much protein markers can add novel insights compared to markers derived from mRNA transcriptomic profiling. Using paired label-free LC-MS/MS and gene expression microarray measurements from primary samples of patients with acute myeloid leukemia (AML), we demonstrate here that while the quantitative proteomic and transcriptomic profiles were highly correlated, in general, the marker panels showing statistically significant expression changes across the disease and healthy groups were profoundly different between protein and mRNA levels. In particular, the proteomic assay enabled unique links to known leukemic processes, which were missed when using the transcriptomic profiling alone, as well as identified additional links to metabolic regulators and chromatin remodelers, such as GPX1, fumarate hydratase, and SET oncogene, which have subsequently been evaluated in independent AML samples. Overall, these results highlighted the complementary and informative view obtained from the quantitative LC-MS/MS approach into the AML deregulated signaling networks.

Designing patient-oriented combination therapies for acute myeloid leukemia based on efficacy/toxicity integration and bipartite network modeling.

Acute myeloid leukemia (AML), a heterogeneous and aggressive blood cancer, does not respond well to single-drug therapy. A combination of drugs is required to effectively treat this disease. Computational models are critical for combination therapy discovery due to the tens of thousands of two-drug combinations, even with approved drugs. While predicting synergistic drugs is the focus of current methods, few consider drug efficacy and potential toxicity, which are crucial for treatment success. To find effective new drug candidates, we constructed a bipartite network using patient-derived tumor samples and drugs. The network is based on drug-response screening and summarizes all treatment response heterogeneity as drug response weights. This bipartite network is then projected onto the drug part, resulting in the drug similarity network. Distinct drug clusters were identified using community detection methods, each targeting different biological processes and pathways as revealed by enrichment and pathway analysis of the drugs' protein targets. Four drugs with the highest efficacy and lowest toxicity from each cluster were selected and tested for drug sensitivity using cell viability assays on various samples. Results show that ruxolitinib-ulixertinib and sapanisertib-LY3009120 are the most effective combinations with the least toxicity and the best synergistic effect on blast cells. These findings lay the foundation for personalized and successful AML therapies, ultimately leading to the development of drug combinations that can be used alongside standard first-line AML treatment.