ArcheD, a residual neural network for prediction of cerebrospinal fluid amyloid-beta from amyloid PET images.

Detection and measurement of amyloid-beta (Aß) in the brain is a key factor for early identification and diagnosis of Alzheimer's disease (AD). We aimed to develop a deep learning model to predict Aß cerebrospinal fluid (CSF) concentration directly from amyloid PET images, independent of tracers, brain reference regions or preselected regions of interest. We used 1870 Aß PET images and CSF measurements to train and validate a convolutional neural network ("ArcheD"). We evaluated the ArcheD performance in relation to episodic memory and the standardized uptake value ratio (SUVR) of cortical Aß. We also compared the brain region's relevance for the model's CSF prediction within clinical-based and biological-based classifications. ArcheD-predicted Aß CSF values correlated with measured Aß CSF values (r = 0.92; q < 0.01), SUVR (rAV45 = -0.64, rFBB = -0.69; q < 0.01) and episodic memory measures (0.33 < r < 0.44; q < 0.01). For both classifications, cerebral white matter significantly contributed to CSF prediction (q < 0.01), specifically in non-symptomatic and early stages of AD. However, in late-stage disease, the brain stem, subcortical areas, cortical lobes, limbic lobe and basal forebrain made more significant contributions (q < 0.01). Considering cortical grey matter separately, the parietal lobe was the strongest predictor of CSF amyloid levels in those with prodromal or early AD, while the temporal lobe played a more crucial role for those with AD. In summary, ArcheD reliably predicted Aß CSF concentration from Aß PET scans, offering potential clinical utility for Aß level determination.

Molecular features encoded in the ctDNA reveal heterogeneity and predict outcome in high-risk aggressive B-cell lymphoma.

Inadequate molecular and clinical stratification of the patients with high-risk diffuse large B-cell lymphoma (DLBCL) is a clinical challenge hampering the establishment of personalized therapeutic options. We studied the translational significance of liquid biopsy in a uniformly treated trial cohort. Pretreatment circulating tumor DNA (ctDNA) revealed hidden clinical and biological heterogeneity, and high ctDNA burden determined increased risk of relapse and death independently of conventional risk factors. Genomic dissection of pretreatment ctDNA revealed translationally relevant phenotypic, molecular, and prognostic information that extended beyond diagnostic tissue biopsies. During therapy, chemorefractory lymphomas exhibited diverging ctDNA kinetics, whereas end-of-therapy negativity for minimal residual disease (MRD) characterized cured patients and resolved clinical enigmas, including false residual PET positivity. Furthermore, we discovered fragmentation disparities in the cell-free DNA that characterize lymphoma-derived ctDNA and, as a proof-of-concept for their clinical application, used machine learning to show that end-of-therapy fragmentation patterns predict outcome. Altogether, we have discovered novel molecular determinants in the liquid biopsy that can noninvasively guide treatment decisions.

DeepIFC: Virtual fluorescent labeling of blood cells in imaging flow cytometry data with deep learning.

Imaging flow cytometry (IFC) combines flow cytometry with microscopy, allowing rapid characterization of cellular and molecular properties via high-throughput single-cell fluorescent imaging. However, fluorescent labeling is costly and time-consuming. We present a computational method called DeepIFC based on the Inception U-Net neural network architecture, able to generate fluorescent marker images and learn morphological features from IFC brightfield and darkfield images. Furthermore, the DeepIFC workflow identifies cell types from the generated fluorescent images and visualizes the single-cell features generated in a 2D space. We demonstrate that rarer cell types are predicted well when a balanced data set is used to train the model, and the model is able to recognize red blood cells not seen during model training as a distinct entity. In summary, DeepIFC allows accurate cell reconstruction, typing and recognition of unseen cell types from brightfield and darkfield images via virtual fluorescent labeling.

Identification of multiplicatively acting modulatory mutational signatures in cancer.

BACKGROUND: A deep understanding of carcinogenesis at the DNA level underpins many advances in cancer prevention and treatment. Mutational signatures provide a breakthrough conceptualisation, as well as an analysis framework, that can be used to build such understanding. They capture somatic mutation patterns and at best identify their causes. Most studies in this context have focused on an inherently additive analysis, e.g. by non-negative matrix factorization, where the mutations within a cancer sample are explained by a linear combination of independent mutational signatures. However, other recent studies show that the mutational signatures exhibit non-additive interactions. RESULTS: We carefully analysed such additive model fits from the PCAWG study cataloguing mutational signatures as well as their activities across thousands of cancers. Our analysis identified systematic and non-random structure of residuals that is left unexplained by the additive model. We used hierarchical clustering to identify cancer subsets with similar residual profiles to show that both systematic mutation count overestimation and underestimation take place. We propose an extension to the additive mutational signature model-multiplicatively acting modulatory processes-and develop a maximum-likelihood framework to identify such modulatory mutational signatures. The augmented model is expressive enough to almost fully remove the observed systematic residual patterns. CONCLUSION: We suggest the modulatory processes biologically relate to sample specific DNA repair propensities with cancer or tissue type specific profiles. Overall, our results identify an interesting direction where to expand signature analysis.

Exome-wide somatic mutation characterization of small bowel adenocarcinoma.

Small bowel adenocarcinoma (SBA) is an aggressive disease with limited treatment options. Despite previous studies, its molecular genetic background has remained somewhat elusive. To comprehensively characterize the mutational landscape of this tumor type, and to identify possible targets of treatment, we conducted the first large exome sequencing study on a population-based set of SBA samples from all three small bowel segments. Archival tissue from 106 primary tumors with appropriate clinical information were available for exome sequencing from a patient series consisting of a majority of confirmed SBA cases diagnosed in Finland between the years 2003-2011. Paired-end exome sequencing was performed using Illumina HiSeq 4000, and OncodriveFML was used to identify driver genes from the exome data. We also defined frequently affected cancer signalling pathways and performed the first extensive allelic imbalance (AI) analysis in SBA. Exome data analysis revealed significantly mutated genes previously linked to SBA (TP53, KRAS, APC, SMAD4, and BRAF), recently reported potential driver genes (SOX9, ATM, and ARID2), as well as novel candidate driver genes, such as ACVR2A, ACVR1B, BRCA2, and SMARCA4. We also identified clear mutation hotspot patterns in ERBB2 and BRAF. No BRAF V600E mutations were observed. Additionally, we present a comprehensive mutation signature analysis of SBA, highlighting established signatures 1A, 6, and 17, as well as U2 which is a previously unvalidated signature. Finally, comparison of the three small bowel segments revealed differences in tumor characteristics. This comprehensive work unveils the mutational landscape and most frequently affected genes and pathways in SBA, providing potential therapeutic targets, and novel and more thorough insights into the genetic background of this tumor type.

Exome and immune cell score analyses reveal great variation within synchronous primary colorectal cancers.

BACKGROUND: Approximately 4% of colorectal cancer (CRC) patients have at least two simultaneous cancers in the colon. Due to the shared environment, these synchronous CRCs (SCRCs) provide a unique setting to study colorectal carcinogenesis. Understanding whether these tumours are genetically similar or distinct is essential when designing therapeutic approaches. METHODS: We performed exome sequencing of 47 primary cancers and corresponding normal samples from 23 patients. Additionally, we carried out a comprehensive mutational signature analysis to assess whether tumours had undergone similar mutational processes and the first immune cell score analysis (IS) of SCRC to analyse the interplay between immune cell invasion and mutation profile in both lesions of an individual. RESULTS: The tumour pairs shared only few mutations, favouring different mutations in known CRC genes and signalling pathways and displayed variation in their signature content. Two tumour pairs had discordant mismatch repair statuses. In majority of the pairs, IS varied between primaries. Differences were not explained by any clinicopathological variable or mutation burden. CONCLUSIONS: The study shows major diversity within SCRCs. Rather than rely on data from one tumour, our study highlights the need to evaluate both tumours of a synchronous pair for optimised targeted therapy.

Integrated data analysis reveals uterine leiomyoma subtypes with distinct driver pathways and biomarkers.

Uterine leiomyomas are common benign smooth muscle tumors that impose a major burden on women's health. Recent sequencing studies have revealed recurrent and mutually exclusive mutations in leiomyomas, suggesting the involvement of molecularly distinct pathways. In this study, we explored transcriptional differences among leiomyomas harboring different genetic drivers, including high mobility group AT-hook 2 (HMGA2) rearrangements, mediator complex subunit 12 (MED12) mutations, biallelic inactivation of fumarate hydratase (FH), and collagen, type IV, alpha 5 and collagen, type IV, alpha 6 (COL4A5-COL4A6) deletions. We also explored the transcriptional consequences of 7q22, 22q, and 1p deletions, aiming to identify possible target genes. We investigated 94 leiomyomas and 60 corresponding myometrial tissues using exon arrays, whole genome sequencing, and SNP arrays. This integrative approach revealed subtype-specific expression changes in key driver pathways, including Wnt/ß-catenin, Prolactin, and insulin-like growth factor (IGF)1 signaling. Leiomyomas with HMGA2 aberrations displayed highly significant up-regulation of the proto-oncogene pleomorphic adenoma gene 1 (PLAG1), suggesting that HMGA2 promotes tumorigenesis through PLAG1 activation. This was supported by the identification of genetic PLAG1 alterations resulting in expression signatures as seen in leiomyomas with HMGA2 aberrations. RAD51 paralog B (RAD51B), the preferential translocation partner of HMGA2, was up-regulated in MED12 mutant lesions, suggesting a role for this gene in the genesis of leiomyomas. FH-deficient leiomyomas were uniquely characterized by activation of nuclear factor erythroid 2-related factor 2 (NRF2) target genes, supporting the hypothesis that accumulation of fumarate leads to activation of the oncogenic transcription factor NRF2. This study emphasizes the need for molecular stratification in leiomyoma research and possibly in clinical practice as well. Further research is needed to determine whether the candidate biomarkers presented herein can provide guidance for managing the millions of patients affected by these lesions.

Towards pan-genome read alignment to improve variation calling.

BACKGROUND: Typical human genome differs from the reference genome at 4-5 million sites. This diversity is increasingly catalogued in repositories such as ExAC/gnomAD, consisting of >15,000 whole-genomes and >126,000 exome sequences from different individuals. Despite this enormous diversity, resequencing data workflows are still based on a single human reference genome. Identification and genotyping of genetic variants is typically carried out on short-read data aligned to a single reference, disregarding the underlying variation. RESULTS: We propose a new unified framework for variant calling with short-read data utilizing a representation of human genetic variation - a pan-genomic reference. We provide a modular pipeline that can be seamlessly incorporated into existing sequencing data analysis workflows. Our tool is open source and available online: https://gitlab.com/dvalenzu/PanVC . CONCLUSIONS: Our experiments show that by replacing a standard human reference with a pan-genomic one we achieve an improvement in single-nucleotide variant calling accuracy and in short indel calling accuracy over the widely adopted Genome Analysis Toolkit (GATK) in difficult genomic regions.

Whole-exome sequencing identifies novel candidate predisposition genes for familial polycythemia vera.

BACKGROUND: Polycythemia vera (PV), characterized by massive production of erythrocytes, is one of the myeloproliferative neoplasms. Most patients carry a somatic gain-of-function mutation in JAK2, c.1849G > T (p.Val617Phe), leading to constitutive activation of JAK-STAT signaling pathway. Familial clustering is also observed occasionally, but high-penetrance predisposition genes to PV have remained unidentified. RESULTS: We studied the predisposition to PV by exome sequencing (three cases) in a Finnish PV family with four patients. The 12 shared variants (maximum allowed minor allele frequency <0.001 in Finnish population in ExAC database) predicted damaging in silico and absent in an additional control set of over 500 Finns were further validated by Sanger sequencing in a fourth affected family member. Three novel predisposition candidate variants were identified: c.1254C > G (p.Phe418Leu) in ZXDC, c.1931C > G (p.Pro644Arg) in ATN1, and c.701G > A (p.Arg234Gln) in LRRC3. We also observed a rare, predicted benign germline variant c.2912C > G (p.Ala971Gly) in BCORL1 in all four patients. Somatic mutations in BCORL1 have been reported in myeloid malignancies. We further screened the variants in eight PV patients in six other Finnish families, but no other carriers were found. CONCLUSIONS: Exome sequencing provides a powerful tool for the identification of novel variants, and understanding the familial predisposition of diseases. This is the first report on Finnish familial PV cases, and we identified three novel candidate variants that may predispose to the disease.

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.

Clonally related uterine leiomyomas are common and display branched tumor evolution.

Uterine leiomyomas are extremely frequent benign smooth muscle tumors often presenting as multiple concurrent lesions and causing symptoms such as abnormal menstrual bleeding, abdominal pain and infertility. While most leiomyomas are believed to arise independently, a few studies have encountered separate lesions harboring identical genetic changes, suggesting a common clonal origin. To investigate the frequency of clonally related leiomyomas, genome-wide tools need to be utilized, and thus little is known about this phenomenon. Using MED12 sequencing and SNP arrays, we searched for clonally related uterine leiomyomas in a set of 103 tumors from 14 consecutive patients who entered hysterectomy owing to symptomatic lesions. Whole-genome sequencing was also utilized to study the genomic architecture of clonally related tumors. This revealed four patients to have two or more tumors that were clonally related, all of which lacked MED12 mutations. Furthermore, some tumors were composed of genetically distinct subclones, indicating a nonlinear, branched model of tumor evolution. DEPDC5 was discovered as a novel tumor suppressor gene playing a role in the progression of uterine leiomyomas. Perhaps counterintuitively­considering Knudson's two-hit hypothesis­a large shared deletion was followed by different truncating DEPDC5 mutations in four clonally related leiomyomas. This study provides insight into the intratumor heterogeneity of these tumors and suggests that a shared clonal origin is a common feature of leiomyomas that do not carry an MED12 mutation. These observations also offer one explanation to the common occurrence of multiple concurrent lesions.

Global metabolomic profiling of uterine leiomyomas.

BACKGROUND: Uterine leiomyomas can be classified into molecularly distinct subtypes according to their genetic triggers: MED12 mutations, HMGA2 upregulation, or inactivation of FH. The aim of this study was to identify metabolites and metabolic pathways that are dysregulated in different subtypes of leiomyomas. METHODS: We performed global metabolomic profiling of 25 uterine leiomyomas and 17 corresponding myometrium specimens using liquid chromatography-tandem mass spectroscopy. RESULTS: A total of 641 metabolites were detected. All leiomyomas displayed reduced homocarnosine and haeme metabolite levels. We identified a clearly distinct metabolomic profile for leiomyomas of the FH subtype, characterised by metabolic alterations in the tricarboxylic acid cycle and pentose phosphate pathways, and increased levels of multiple lipids and amino acids. Several metabolites were uniquely elevated in leiomyomas of the FH subtype, including N6-succinyladenosine and argininosuccinate, serving as potential biomarkers for FH deficiency. In contrast, leiomyomas of the MED12 subtype displayed reduced levels of vitamin A, multiple membrane lipids and amino acids, and dysregulation of vitamin C metabolism, a finding which was also compatible with gene expression data. CONCLUSIONS: The study reveals the metabolomic heterogeneity of leiomyomas and provides the requisite framework for strategies designed to target metabolic alterations promoting the growth of these prevalent tumours.

MED12 mutations and FH inactivation are mutually exclusive in uterine leiomyomas.

BACKGROUND: Uterine leiomyomas from hereditary leiomyomatosis and renal cell cancer (HLRCC) patients are driven by fumarate hydratase (FH) inactivation or occasionally by mediator complex subunit 12 (MED12) mutations. The aim of this study was to analyse whether MED12 mutations and FH inactivation are mutually exclusive and to determine the contribution of MED12 mutations on HLRCC patients' myomagenesis. METHODS: MED12 exons 1 and 2 mutation screening and 2SC immunohistochemistry indicative for FH deficiency was performed on a comprehensive series of HLRCC patients' (122 specimens) and sporadic (66 specimens) tumours. Gene expression analysis was performed using Affymetrix GeneChip Human Exon Arrays (Affymetrix, Santa Clara, CA, USA). RESULTS: Nine tumours from HLRCC patients harboured a somatic MED12 mutation and were negative for 2SC immunohistochemistry. All remaining successfully analysed lesions (107/116) were deficient for FH. Of sporadic tumours, 35/64 were MED12 mutation positive and none displayed a FH defect. In global gene expression analysis FH-deficient tumours clustered together, whereas HLRCC patients' MED12 mutation-positive tumours clustered together with sporadic MED12 mutation-positive tumours. CONCLUSIONS: Somatic MED12 mutations and biallelic FH inactivation are mutually exclusive in both HLRCC syndrome-associated and sporadic uterine leiomyomas. The great majority of HLRCC patients' uterine leiomyomas are caused by FH inactivation, but incidental tumours driven by somatic MED12 mutations also occur. These MED12 mutation-positive tumours display similar expressional profiles with their sporadic counterparts and are clearly separate from FH-deficient tumours.

Characterization of uterine leiomyomas by whole-genome sequencing.

BACKGROUND: Uterine leiomyomas are benign but affect the health of millions of women. A better understanding of the molecular mechanisms involved may provide clues to the prevention and treatment of these lesions. METHODS: We performed whole-genome sequencing and gene-expression profiling of 38 uterine leiomyomas and the corresponding myometrium from 30 women. RESULTS: Identical variants observed in some separate tumor nodules suggested that these nodules have a common origin. Complex chromosomal rearrangements resembling chromothripsis were a common feature of leiomyomas. These rearrangements are best explained by a single event of multiple chromosomal breaks and random reassembly. The rearrangements created tissue-specific changes consistent with a role in the initiation of leiomyoma, such as translocations of the HMGA2 and RAD51B loci and aberrations at the COL4A5-COL4A6 locus, and occurred in the presence of normal TP53 alleles. In some cases, separate events had occurred more than once in single tumor-cell lineages. CONCLUSIONS: Chromosome shattering and reassembly resembling chromothripsis (a single genomic event that results in focal losses and rearrangements in multiple genomic regions) is a major cause of chromosomal abnormalities in uterine leiomyomas; we propose that tumorigenesis occurs when tissue-specific tumor-promoting changes are formed through these events. Chromothripsis has previously been associated with aggressive cancer; its common occurrence in leiomyomas suggests that it also has a role in the genesis and progression of benign tumors. We observed that multiple separate tumors could be seeded from a single lineage of uterine leiomyoma cells. (Funded by the Academy of Finland Center of Excellence program and others.).

Eleven candidate susceptibility genes for common familial colorectal cancer.

Hereditary factors are presumed to play a role in one third of colorectal cancer (CRC) cases. However, in the majority of familial CRC cases the genetic basis of predisposition remains unexplained. This is particularly true for families with few affected individuals. To identify susceptibility genes for this common phenotype, we examined familial cases derived from a consecutive series of 1514 Finnish CRC patients. Ninety-six familial CRC patients with no previous diagnosis of a hereditary CRC syndrome were included in the analysis. Eighty-six patients had one affected first-degree relative, and ten patients had two or more. Exome sequencing was utilized to search for genes harboring putative loss-of-function variants, because such alterations are likely candidates for disease-causing mutations. Eleven genes with rare truncating variants in two or three familial CRC cases were identified: UACA, SFXN4, TWSG1, PSPH, NUDT7, ZNF490, PRSS37, CCDC18, PRADC1, MRPL3, and AKR1C4. Loss of heterozygosity was examined in all respective cancer samples, and was detected in seven occasions involving four of the candidate genes. In all seven occasions the wild-type allele was lost (P = 0.0078) providing additional evidence that these eleven genes are likely to include true culprits. The study provides a set of candidate predisposition genes which may explain a subset of common familial CRC. Additional genetic validation in other populations is required to provide firm evidence for causality, as well as to characterize the natural history of the respective phenotypes.

Comparative genome-scale reconstruction of gapless metabolic networks for present and ancestral species.

We introduce a novel computational approach, CoReCo, for comparative metabolic reconstruction and provide genome-scale metabolic network models for 49 important fungal species. Leveraging on the exponential growth in sequenced genome availability, our method reconstructs genome-scale gapless metabolic networks simultaneously for a large number of species by integrating sequence data in a probabilistic framework. High reconstruction accuracy is demonstrated by comparisons to the well-curated Saccharomyces cerevisiae consensus model and large-scale knock-out experiments. Our comparative approach is particularly useful in scenarios where the quality of available sequence data is lacking, and when reconstructing evolutionary distant species. Moreover, the reconstructed networks are fully carbon mapped, allowing their use in 13C flux analysis. We demonstrate the functionality and usability of the reconstructed fungal models with computational steady-state biomass production experiment, as these fungi include some of the most important production organisms in industrial biotechnology. In contrast to many existing reconstruction techniques, only minimal manual effort is required before the reconstructed models are usable in flux balance experiments. CoReCo is available at http://esaskar.github.io/CoReCo/.

Exome sequencing reveals frequent inactivating mutations in ARID1A, ARID1B, ARID2 and ARID4A in microsatellite unstable colorectal cancer.

ARID1A has been identified as a novel tumor suppressor gene in ovarian cancer and subsequently in various other tumor types. ARID1A belongs to the ARID domain containing gene family, which comprises of 15 genes involved, for example, in transcriptional regulation, proliferation and chromatin remodeling. In this study, we used exome sequencing data to analyze the mutation frequency of all the ARID domain containing genes in 25 microsatellite unstable (MSI) colorectal cancers (CRCs) as a first systematic effort to characterize the mutation pattern of the whole ARID gene family. Genes which fulfilled the selection criteria in this discovery set (mutations in at least 4/25 [16%] samples, including at least one nonsense or splice site mutation) were chosen for further analysis in an independent validation set of 21 MSI CRCs. We found that in addition to ARID1A, which was mutated in 39% of the tumors (18/46), also ARID1B (13%, 6/46), ARID2 (13%, 6/46) and ARID4A (20%, 9/46) were frequently mutated. In all these genes, the mutations were distributed along the entire length of the gene, thus distinguishing them from typical MSI target genes previously described. Our results indicate that in addition to ARID1A, other members of the ARID gene family may play a role in MSI CRC.

Identification of candidate oncogenes in human colorectal cancers with microsatellite instability.

Microsatellite instability can be found in approximately 15% of all colorectal cancers. To detect new oncogenes we sequenced the exomes of 25 colorectal tumors and respective healthy colon tissue. Potential mutation hot spots were confirmed in 15 genes; ADAR, DCAF12L2, GLT1D1, ITGA7, MAP1B, MRGPRX4, PSRC1, RANBP2, RPS6KL1, SNCAIP, TCEAL6, TUBB6, WBP5, VEGFB, and ZBTB2; these were validated in 86 tumors with microsatellite instability. ZBTB2, RANBP2, and PSRC1 also were found to contain hot spot mutations in the validation set. The form of ZBTB2 associated with colorectal cancer increased cell proliferation. The mutation hot spots might be used to develop personalized tumor profiling and therapy.

Single-cell transcriptomes identify patient-tailored therapies for selective co-inhibition of cancer clones.

Intratumoral cellular heterogeneity necessitates multi-targeting therapies for improved clinical benefits in advanced malignancies. However, systematic identification of patient-specific treatments that selectively co-inhibit cancerous cell populations poses a combinatorial challenge, since the number of possible drug-dose combinations vastly exceeds what could be tested in patient cells. Here, we describe a machine learning approach, scTherapy, which leverages single-cell transcriptomic profiles to prioritize multi-targeting treatment options for individual patients with hematological cancers or solid tumors. Patient-specific treatments reveal a wide spectrum of co-inhibitors of multiple biological pathways predicted for primary cells from heterogenous cohorts of patients with acute myeloid leukemia and high-grade serous ovarian carcinoma, each with unique resistance patterns and synergy mechanisms. Experimental validations confirm that 96% of the multi-targeting treatments exhibit selective efficacy or synergy, and 83% demonstrate low toxicity to normal cells, highlighting their potential for therapeutic efficacy and safety. In a pan-cancer analysis across five cancer types, 25% of the predicted treatments are shared among the patients of the same tumor type, while 19% of the treatments are patient-specific. Our approach provides a widely-applicable strategy to identify personalized treatment regimens that selectively co-inhibit malignant cells and avoid inhibition of non-cancerous cells, thereby increasing their likelihood for clinical success.