Advertisement
Browse Subject Areas
?

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

  • Loading metrics

Open Access

Peer-reviewed

Research Article

Higher order synthetic lethals are keys to minimize cancer treatment effects on non-tumor cells

  • Mehdi Dehghan Manshadi,

    Roles Formal analysis, Methodology, Software, Visualization, Writing – original draft

    Affiliations Department of Oncology-Pathology, Karolinska Institutet, Stockholm, Sweden, Algorithmic Dynamics Lab, Center for Molecular Medicine, Karolinska University Hospital, Stockholm, Sweden, Department of Chemical Engineering, School of Chemical, Petroleum and Gas Engineering, Shiraz University, Shiraz, Iran

    ⨯
  • Payam Setoodeh ,

    Roles Conceptualization, Methodology, Project administration, Resources, Supervision, Writing – review & editing

    payamst@shirazu.ac.ir, setoodep@mcmaster.ca (PS); rajaba@mcmaster.ca (ARR); zare@uthscsa.edu (HZ)

    Affiliations Department of Chemical Engineering, School of Chemical, Petroleum and Gas Engineering, Shiraz University, Shiraz, Iran, W Booth School of Engineering Practice and Technology, McMaster University, Hamilton, Ontario, Canada

    ⨯
  • Amin Ramezani,

    Roles Conceptualization, Project administration, Writing – review & editing

    Affiliations Shiraz Institute for Cancer Research, School of Medicine, Shiraz University of Medical Sciences, Shiraz, Iran, Department of Medical Biotechnology, School of Advanced Medical Sciences and Technologies, Shiraz University of Medical Sciences, Shiraz, Iran

    ⨯
  • Amin Reza Rajabzadeh ,

    Roles Project administration, Writing – review & editing

    payamst@shirazu.ac.ir, setoodep@mcmaster.ca (PS); rajaba@mcmaster.ca (ARR); zare@uthscsa.edu (HZ)

    ☯ These authors contributed equally as the co-senior authors.

    Affiliations W Booth School of Engineering Practice and Technology, McMaster University, Hamilton, Ontario, Canada, School of Biomedical Engineering, McMaster University, Hamilton, Ontario, Canada

    ⨯
  • Habil Zare

    Roles Methodology, Project administration, Supervision, Writing – review & editing

    payamst@shirazu.ac.ir, setoodep@mcmaster.ca (PS); rajaba@mcmaster.ca (ARR); zare@uthscsa.edu (HZ)

    ☯ These authors contributed equally as the co-senior authors.

    Affiliation Haba Businesses, San Antonio, Texas, United States of America

    ⨯

Abstract

Metabolic rewiring in cancer cells facilitates the supply of essential precursors for uncontrolled tumor growth. Exploring these cancer-specific metabolic alterations offers potential selective therapeutic strategies. However, targeting a single essential gene in cancer treatment often faces challenges, including resistance, lack of targetable oncogenes, and potential harm to non-tumor cells. Targeting multiple genes has been proposed as a solution to overcome these issues, e.g., a synthetic lethal (SL) set, defined as a minimal combination of non-lethal genetic perturbations that lead to cell death. This study theoretically examined the potential of SL sets to identify selective drug targets across 13 cancer types and the corresponding non-tumor tissues, utilizing context-specific genome-scale metabolic models. To ensure the minimization of therapeutic side effects, this work introduced the concept of strictly-selective drug targets (SSDTs) and the lack of harmful effects of the identified targets in all 13 different non-tumor tissues was meticulously verified. Accordingly, for 13 types of cancers, over 500 SSDTs were identified, predominantly including higher-order SL sets with more than two targets in each set. Interestingly, for specific cancers where single essential or SL genes could not provide viable therapeutic solutions, SSDTs were provided by higher-order SL sets. Thus, for the first time, this study demonstrates that leveraging higher-order SL sets may offer promising strictly selective therapeutic solutions. Furthermore, nine quadruple SSDTs were identified, which commonly target five different cancers without harming any of the 13 non-tumor tissues. Further experimental validation of these findings is essential to identify the most promising treatment candidates for future clinical studies/applications.

Citation: Dehghan Manshadi M, Setoodeh P, Ramezani A, Rajabzadeh AR, Zare H (2026) Higher order synthetic lethals are keys to minimize cancer treatment effects on non-tumor cells. PLoS One 21(4): e0342199. https://doi.org/10.1371/journal.pone.0342199

Editor: Bashir Sajo Mienda, Federal University Dutse, NIGERIA

Received: February 13, 2025; Accepted: January 19, 2026; Published: April 15, 2026

Copyright: © 2026 Dehghan Manshadi et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

Data Availability: All data, code, and supplementary information related to this work are available at https://github.com/CSBLaboratory/CancerSSDTs.

Funding: This work is supported by NIH–NIA (grant numbers: R01AG057896, 1RF1AG063507, R01AG068293, 1R01AG0665241A,1R01AG065301, P30 AG066546, and 1R21AG087907) and NIH–NINDS (grant numbers: RF1NS112391, U19NS115388, and R015R01NS017950-39). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Competing interests: The authors have declared that no competing interests exist.

Introduction

Metabolic rewiring leads to distinct alterations in flux distributions across various pathways in cancer cells. Otto Warburg was the first to highlight these metabolic variations, demonstrating that cancer cells preferentially rely on aerobic glycolysis rather than oxidative phosphorylation [1]. This phenomenon, known as the Warburg effect, illustrates how these metabolic alterations in cancer cells support the increased demand for fatty acids, nucleotides, and amino acids [2]. Extensive research has focused on aerobic glycolysis to better understand cancer cell metabolism [3–13].

In addition to the Warburg effect, numerous other metabolic alterations have been reported across different cancer types. One such alteration is the upregulation of the glutaminolysis pathway, which, together with enhanced glycolysis, supplies both energy and biosynthetic precursors to sustain elevated proliferation rates [14–16]. Furthermore, the literature indicates the deregulation of other pathways, such as the lipid synthesis pathway [17], the branched-chain amino acid metabolism, the serine synthesis pathway [18], and the pentose phosphate pathway [19]. Metabolic rewiring in cancer cells also presents unique opportunities for immune evasion [20], apoptosis disruption [21], and the utilization of metabolic byproducts such as ammonia, ketone bodies, acetate, and lactate [18].

Characterization and identification of the mentioned cancer-specific metabolic alterations pave the way for developing selective therapeutic strategies that exclusively affect cancer cells without harming non-tumor tissues [22–24]. Genome-scale metabolic models (GEMs), which represent comprehensive mathematical reconstructions of cellular metabolism, have emerged as powerful tools for studying metabolic rewiring at the systems level [25–27]. Consequently, various computational approaches have been developed to use these models for identifying metabolic cancer drug targets [28,29]. Folger et al. [30] provided one of the earliest studies in the field of drug target identification using GEMs. The authors provided a unified model using expression data from different cancers and found 52 drug targets, 21 of which corresponded to experimentally validated or known anticancer drugs. To generate context- or patient-specific GEMs, a number of approaches have been developed and applied to identify different cancer therapeutics [31]. Yizhak et al. [32] constructed more than 280 healthy and cancerous models using PRIME (personalized reconstruction of metabolic models). They applied this set of cell-specific models to predict selective drug targets, experimentally validating the top predicted target, MLYCD, and investigating the depletion effects of this gene. In another study, Gatto et al. [33] used the INIT algorithm [34] to develop GEMs for clear cell renal cell carcinoma (ccRCC) and successfully identified five metabolic genes as selective drug targets, predicted to be dispensable in non-tumor cell metabolism. Barrena et al. [35] extended the gMCS approach by integrating linear regulatory pathways with metabolic models using Human1 and regulatory databases. They identified new essential genes and synthetic lethal (SL) pairs in cancer cell lines. Larsson et al. [36] used a generic GEM to identify five selective essential genes for glioblastoma, with in vitro or in vivo evidence of the essentiality of 4 out of the 5 being reported in the literature. Pacheco et al. [37] introduced the rFASTCORMICS algorithm and constructed 10,005 GEMs for 13 different types of cancers. The authors performed a single gene essentiality analysis to identify selective drug targets and used colorectal cancer as a successful experimental test case for drug repurposing of mimosine, ketoconazole, and naftifin.

Despite valuable single-drug targets being identified using the mentioned approaches, based on the literature, targeting only one gene presents several challenges [38]. The survival dependency of a tumor on an oncogene or oncogenic pathway is a phenomenon known as oncogene addiction. Currently, this phenomenon forms the basis of most genotype-targeted cancer therapeutics. However, gain-of-function oncogenes are not targetable in all tumors, resulting in common resistance to the administered therapy [39]. Furthermore, identifying unmarked oncogenes, the cancer-specific driver genes with no evidence of genetic alterations, is challenging. To address these limitations, the concept of synthetic lethality has been proposed [39–41]. Synthetic lethality describes a pair of genes for which individual perturbations are non-lethal, whereas their simultaneous disruption results in cell death. This principle offers the potential to selectively target cancer cells by exploiting tumor-specific genetic or metabolic vulnerabilities [42]. Moreover, SL sets provide opportunities to target cancers with pathway alterations that are otherwise considered undruggable [43]. Although the capabilities of SLs for providing selective drug targets are mentioned in the literature, relatively few studies have systematically explored their application. In this context, Frezza et al. [44] investigated fumarate hydratase Fh1-deficient kidney cells. The researchers constructed context-specific genome-scale metabolic models (GEMs) for cell lines with and without Fh1, using gene expression data. They predicted gene knockouts that were lethal to Fh1-deficient cells but did not affect non-tumor cells. The analysis highlighted reactions—primarily in the heme metabolism pathway—as synthetically lethal with Fh1 deletion. In another study, Zhang et al. [45] performed an exhaustive search to identify 44 double and 95 triple selective drug targets for hepatocellular carcinoma (HCC) using a generic GEM from Argen et al. [46].

In addition to mechanistic GEM-based strategies, several computational and machine-learning approaches have been developed to predict SL interactions from large-scale genomic and transcriptomic data. These include network-based models, matrix-completion frameworks, and emerging deep-learning architectures [47]. While such methods offer complementary, data-driven perspectives, their predictive performance remains variable, especially for multiple-gene intervention, and further methodological refinement is needed before they can reliably guide SL discovery on their own [48].

Although synthetic lethality is considered a more robust strategy for identifying selective drug targets compared to single-gene targeting, to the best of our knowledge, no systematic study has yet demonstrated this advantage. Therefore, this work aims to identify SL gene sets predicted by analyses of context-specific genome-scale metabolic models to determine the capabilities of the concept of synthetic lethality for the identification of strictly-selective drug targets. Here, the term “strictly-selective” refers to targets that affect cancer cells while minimizing damage not only to the corresponding non-tumor tissue but also to non-tumor tissues across different organs. To achieve this objective, context-specific models were constructed for 13 different pairs of cancerous and non-tumor tissues using the rFASTCORMICS algorithm. The names and abbreviations of the studied cancers are listed in Table 1.

thumbnail
Download:
Table 1. TCGA abbreviations of the cancers studied in this paper.

https://doi.org/10.1371/journal.pone.0342199.t001

For these models, essential genes, SL pairs, and higher-order SL sets (comprising up to four genes), were identified. Subsequently, potential strictly-selective drug targets were determined. The results showed that when a few or even no essential genes can be targeted due to causing side effects for different non-tumor cells in the human body, higher-order SL sets represent a key strategy for identifying strictly-selective drug targets that discriminate solely against cancerous tissue(s). Accordingly, in addition to the identification of several strictly-selective potential drug targets for 13 cancers, for the first time, this study successfully showed the capabilities of the concept of synthetic lethality for providing strictly selective drug targets for cancers applying genome-scale metabolic modeling and in-silico experimentations/analyses.

Methods

Reconstruction of context-specific models

In this study, 26 context-specific models, consisting of 13 generic control models and 13 generic cancer models, were reconstructed using rFASTCORMICS [37], a promising tool for generating accurate models capable of predicting the outcomes of gene knockout strategies. The Recon 2.04 [49] genome-scale reconstruction (RRID:SCR_006345) was chosen as the reference model, and the RNA-seq data (GSE62944) [50] were retrieved from the TCGA Research Network (http://cancergenome.nih.gov/). The assignment of samples to cancerous and non-tumor tissues was based on their pathology information available from TCGA. This process could not be randomized. Also, power analysis was not applicable here because we used all existing data.

While arbitrary thresholds for RNA-seq data can substantially influence model precision in classifying expressed and non-expressed genes [51], rFASTCORMICS adopts an approach proposed by Hart et al. [52] to determine active genes. Specifically, rFASTCORMICS generates a density plot for each sample based on its log2-transformed FPKM (Fragments per Kilobase of transcript per Million mapped reads) values. Subsequently, a Gaussian curve is fitted to the right half of the main peak (expression curve). The mean value of this curve is utilized as the expression threshold. Then, the obtained curve is subtracted from the density curve, and another Gaussian curve is fitted to this part of the density curve (inexpression curve).

rFASTCORMICS uses the mean value (μ) and standard deviation (σ) of the right-hand Gaussian curve to convert log2-transformed FPKM values to the zFPKM values.

(1)

The mean of the inexpression curve was chosen as a threshold if it was above −3 zFPKM; otherwise, −3 was taken, and genes with lower zFPKM values were considered inactive with a score of −1. Conversely, 0 zFPKM was chosen as the expression threshold, and genes with higher zFPKM values were considered active with a score of +1. Genes with zFPKM values between these two thresholds were considered to have an unknown expression status with a score of 0. Finally, using gene-protein-reaction (GPR) rules, expression scores for genes were mapped to corresponding reactions, and bounds for inactive reactions were set to zero.

Extracting SL sets

Extracting SL gene pairs from reconstructed models is not computationally challenging. However, extending the synthetic lethality principle to gene sets with more than two genes and attempting to identify higher-order SL sets (HOSLs) presents significant computational challenges. As the number of genes in an SL set increases, the corresponding search space grows exponentially, making the search process extremely time-consuming. Therefore, an algorithm is required to explore this expanding search space efficiently. In this study, Rapid-SL [53] was applied to identify HOSLs up to quadruples (four genes in a set) for 13 cancerous models. Alternative methods for determining all HOSLs, include gMCS [54] and Fast-SL [55], although these may not be as computationally efficient as Rapid-SL in practice.

Strict selectivity criteria

The primary goal of this work was to identify strictly-selective potential drug targets. In this study, strictly-selective drug targets refer to essential genes or SL gene sets. These targets aim to reduce the growth rate of cancer cells without adversely affecting non-tumor tissues across other organs. To mathematically describe these criteria, certain thresholds were used. Here, a gene (or gene set) was considered lethal (or synthetic lethal) if its knockout reduced the growth rate of cancer cells by over 50% [37]. Furthermore, not affecting non-tumor tissues was defined as causing less than a 10% [37] reduction in the growth rate of non-tumor tissues when the gene (or gene set) was knocked out from the corresponding models. Fig 1 shows a schematic overview of the procedure used in this work.

thumbnail
Download:
Fig 1. Overview of the workflow followed in this study.

https://doi.org/10.1371/journal.pone.0342199.g001

Results

This study identified essential genes and SL gene sets for 13 context-specific genome-scale models of different cancers. Additionally, besides identifying SL genes, this study successfully identified triple and quadruple SL gene sets, referred to as higher-order SL sets. The list of strictly-selective drug targets obtained in this work is provided in the S1 File. We did not apply druggability filtering when identifying SSDTs. Some strictly selective vulnerabilities involve genes that are not currently tractable by small-molecule inhibitors (e.g., RPE). We chose to report the full spectrum of model-derived vulnerabilities because non-druggable genes may still be useful for biomarker development, pathway dissection, synthetic-lethality sensitization, or future therapeutic strategies that extend beyond conventional small-molecule targeting. Therefore, the presence of non-druggable genes reflects our intention to provide an unbiased and comprehensive map of strictly selective metabolic vulnerabilities.

Cases with essential genes as strictly-selective solutions

Here, the reported results are explained using colon cancer as an example. The number of essential genes and SL sets obtained for the COAD GEM is reported in Table 2. This table illustrates how the number of solutions decreased when their selectivity was tested using the GEM of the non-tumor tissue of the colon.

thumbnail
Download:
Table 2. Number of essential and SL solutions identified for colon cancer.

https://doi.org/10.1371/journal.pone.0342199.t002

Although the solutions obtained in this step are potentially selective for cancerous colon tissue cells, there remains a possibility of affecting other non-tumor tissues. Therefore, the effect of the remaining solutions on other non-tumor tissues was examined step by step by considering the selectivity criteria for other non-tumor cells, such as breast, lung, etc. By considering each additional non-tumor tissue, some potential solutions may be found to be non-selective and omitted from the list of potential selective drug targets. When considering all 13 non-tumor tissues, many candidate solutions were excluded. As a result, SLC38A3, a member of the transport of inorganic cations/anions and amino acids/oligopeptides pathway, has been identified as a single strictly-selective drug target. In comparison to SQLE proposed by Pacheco et al. [37], which has side effects on other non-tumor tissues, such as the liver or kidney, SLC38A3 is expected to have minimal side effects on all 13 non-tumor tissues. Asparagine (DB00174) is one of the compounds reported to inhibit SLC38A3 and may serve as a candidate for validating this strictly-selective drug target.

Although no double strictly-selective SL set was found, two three-membered SL sets were identified for COAD. The first set includes RPE, HSD11B2, and G6PD. Two of these genes also appear in the second SL set, which includes RPE, HSD11B2, and PGLS. These genes are functionally related. Specifically, RPE, G6PD, and PGLS contribute to the pentose phosphate pathway and the gene HSD11B2 is related to the glucocorticoid biosynthesis pathway. Various drugs are reported to have effects on G6PD, PGLS, and HSD11B2 (see S2 File); however, RPE is not listed as a target for any drugs on DrugBank [56]. Because all genes in the set of RPE, G6PD, and HSD11B2 are reported to have fold changes greater than 1 [57], we propose this lethal set as a suitable candidate for further investigations.

By increasing the cardinality of the SL sets to four, the number of strictly-selective drug targets increases to nine sets. The genes included in these nine sets are depicted in Fig 2. The characteristics of these genes are discussed in the Discussion section.

thumbnail
Download:
Fig 2. The nine mutual quadruple SL sets among COAD, BRCA, HNSC, UCEC, and STAD.

All sets share three genes in common: PFKL, TALDO1, and.

https://doi.org/10.1371/journal.pone.0342199.g002

To examine how protecting additional non-tumor tissues affects the number of potential solutions, the number of single essential, double SL, triple SL, and quadruple SL solutions that remained selective at each step is presented in Fig 3.

thumbnail
Download:
Fig 3. Number of potential selective drug targets identified when different non-tumor tissues were considered protected and COAD was considered the target.

The non-tumor tissues considered protected in each column are marked in green below the chart.

https://doi.org/10.1371/journal.pone.0342199.g003

In Fig 3, from left to right, as the number of tissues considered protected increases, the number of selective drug targets decreases. Considering only six non-tumor tissues (including the non-tumor tissue of the colon itself), all double SL sets were disqualified, leaving only one essential gene as a selective target. However, 11 higher-order SL sets (nine quadruple SLs and two triple SLs) remained as strictly-selective solutions.

As an example of these higher-order SLs, the simultaneous knockout of PFKL, TALDO1, PGD, and CMAS was identified as deleterious for COAD cells, with potentially minimal side effects on the 13 non-tumor tissues. For other cancers, the same figures were generated and reported in Figs 4 and Fig 5.

thumbnail
Download:
Fig 4. Number of potential selective drug targets identified when different non-tumor tissues were considered protected and BRCA, LUAD, LUSC, and UCEC were considered the targets.

https://doi.org/10.1371/journal.pone.0342199.g004

thumbnail
Download:
Fig 5. Number of potential selective drug targets identified when different non-tumor tissues were considered protected and STAD, KIRC, and PRAD were considered the targets.

https://doi.org/10.1371/journal.pone.0342199.g005

For these eight cases (shown in Figs 3, 4, 5), no SL set (sets with two targets) was found to meet the strictly-selectivity criteria, while higher-order SL sets, especially quadruple SLs, emerged as the dominant source of potential solutions.

Cases without any essential genes as strictly-selective solutions

After obtaining strictly-selective drug targets for all 13 cancers, it was revealed that no single essential drug target was found for five cases: LIHC, HNSC, KICH, THCA, and KIRP. This suggests that essential genes of these cancers are also essential for at least one non-tumor tissue. Interestingly, for all of these cases, SLs, and more significantly, higher-order SLs, offer effective strategies. The number of selective solutions at each step for these five cases is shown in Fig 6.

thumbnail
Download:
Fig 6. Number of potential selective drug targets identified when different non-tumor tissues were considered protected and LIHC, HNSC, KICH, THCA, and KIRP were considered the targets.

https://doi.org/10.1371/journal.pone.0342199.g006

Exploration of the identified solutions across different cancers

In this section, different aspects of the obtained solutions are investigated. By considering all the obtained solutions, the most frequently occurring gene, as shown in Fig 7, is Phosphogluconate Dehydrogenase (PGD), followed by Dihydrolipoyl transacetylase (DLAT) and solute carrier family 25 member 20 (SLC25A20). Other important genes, such as Transaldolase 1 (TALDO1) and 6-phosphofructokinase liver type (PFKL), were targeted in several solutions.

thumbnail
Download:
Fig 7. Frequency of different genes across all identified solutions.

This figure shows how many times each gene was identified as a target across all single essential and SL solutions.

https://doi.org/10.1371/journal.pone.0342199.g007

The relationship between different genes in the obtained solutions is visualized in Fig 8 based on their co-occurrence in each SL set. Because various solutions were identified for LIHC and KIRC, their graph is not presented here (see S3 File).

thumbnail
Download:
Fig 8. Relationships among genes in the identified strictly-selective solutions.

Single essential genes are shown in red, double SLs are shown in purple and connected by wavy lines, triple SLs are shown in green and connected by dashed lines, and quadruple SLs are shown in blue and connected by solid lines.

https://doi.org/10.1371/journal.pone.0342199.g008

According to Fig 8, COAD, BRCA, HNSC, UCEC, and STAD share a group of nine quadruple SL sets that consist of a core of genes, including PFKL, TALDO1, and PGD. In each set, the fourth gene provides the decisive perturbation and reduces the growth rate of the mutant to below 50% of the corresponding cancer cell.

Discussion

In this work, strictly-selective potential drug targets were identified as those that are lethal to cancer cells while remaining inactive not only in the corresponding healthy tissue but also across other non-tumor tissues in the human body. Although for selective targeting of cancer cells, it is a common practice to necessarily target a cancer-specific mutant gene alongside a non-essential gene [39,40], this work did not limit the identification of SL pairs to cancer-specific mutations. Through this unbiased approach, a wider range of solutions was identified, which does include solutions based on the cancer-specific mutations. Note that the provided results are solely based on in silico experiments, and further examinations are crucial for each specific potential case, especially because some of the obtained solutions, such as SLC38A3, RPE, HSD11B2, or PGLS identified as a target for different cancers in this study, have been reported as deleterious in mice [58].

To obtain the strictly-selective drug targets, 26 context-specific genome-scale metabolic models were reconstructed for non-tumor and cancerous cells of 13 tissues using the rFASTCORMICS algorithm. Afterward, synthetic lethality analysis on the 13 cancerous models was performed using Rapid-SL to identify double, triple, and quadruple SL gene sets, as well as single essential genes. As a result, desired drug targets were identified for each cancer by checking the strictly-selectivity criteria on each essential gene or SL gene set using the 13 non-tumor GEMs.

The superiority of higher-order SL sets to provide potential solutions was demonstrated by analyzing the obtained solutions. It also revealed that no strictly-selective solution could be provided by single essential genes for LIHC, HNSC, KICH, THCA, and KIRP, and only synthetic and higher-order SL sets were able to offer qualified solutions for these cases. Furthermore, nine quadruple SL sets were identified as strictly-selective for COAD, BRCA, HNSC, UCEC, and STAD. These nine sets include three shared genes: PFKL (phosphofructokinase), TALDO1 (transaldolase), and PGD (phosphogluconate dehydrogenase), and nine different genes for each set. The three shared genes are among the most frequently identified genes in the solutions and are involved in both glycolysis and the pentose phosphate pathway, both of which are highly implicated in the development of cancers. Due to the crucial role of pentose phosphate pathway overexpression in supporting the anabolic needs and addressing oxidative stress in glycolytic cancer cells [59,60], simultaneously targeting both pathways is expected to be effective against cancers.

Phosphofructokinase is a key enzyme in glycolysis and gluconeogenesis, and its inhibition has been reported to be effective in suppressing colorectal cancer [61]. Additionally, it is known to halt the progression of hepatocellular carcinoma [62] and manage lung cancer [63]. Transaldolase is an enzyme related to the pentose phosphate pathway and has been associated with the development of cancer [64]. This enzyme plays a vital role in the pentose phosphate pathway within fast-dividing cancer cells, serving as a crucial mediator that enables glycolytic cancer cells to meet their anabolic needs and effectively cope with oxidative stress [60]. Additionally, this enzyme is associated with bladder [65], liver [66,67], and breast [68] cancer. The association between higher transaldolase expression and decreased responsiveness to HER2 inhibitors in breast cancer patients suggests a potential role of transaldolase in cancer drug resistance. Therefore, suppressing transaldolase as a drug target increases the susceptibility of HER2-amplified cell lines to HER2 inhibition [69]. Phosphogluconate dehydrogenase is one of the pentose phosphate pathway enzymes that has recently gained attention due to its crucial role in the development of tumors and maintaining cellular redox balance [70]. Phosphogluconate dehydrogenase is frequently overexpressed in various types of cancer, promoting the ability to grow and spread in different cancer cells, such as breast [71,72], ovarian [73], and lung [74]. Additionally, overexpression of phosphogluconate dehydrogenase is related to the development of resistance to radiotherapy [75] and chemotherapy [70,76,77] in different cancers.

Besides the three shared genes, nine distinct genes exist in each of the nine identified strictly-selective potential drug targets. These genes are related to five pathways: CTSA, GLB1, and NEU1 are related to the sphingolipid metabolism pathway, which is altered in the development and progression of different cancers, especially colon cancer [78–83]. SLC35A1, SLC35A2, and SLC35D2 are related to the solute carrier family involved in Golgi transport. Furthermore, the SLC family has been reported to have a significant role in anticancer drug resistance [84]. SLC35A2 has been shown to be upregulated in many cancers [85,86], including colon cancer. CMAS, or cytidine monophosphate N-acetylneuraminic acid synthetase, is related to amino-sugar metabolism. Knockdown of CMAS has anticancer activity in triple-negative breast cancer [87]. GALNS, or N-acetylgalactosamine-6-sulfatase, is related to keratan sulfate biosynthesis and is reported to have higher expression in many cancers [88]. ENTPD4, or ectonucleoside triphosphate diphosphohydrolase 4, is known to have hydrolase activity and specifically functions as a UDP phosphatase, being related to nucleotide metabolism. However, ENTPD4 is a novel gene that has been investigated in a few research studies. Recently, a study showed that ENTPD4 is overexpressed in gastric cancer tissues [89]. Hence, its downregulation or knockout may provide a path for competing cancers.

Most experimentally validated SL interactions reported in the literature involve regulatory or signaling genes [90], reflecting the fact that these areas have been the primary focus of functional genomic screens. In contrast, metabolic synthetic lethality, particularly higher-order combinations involving three or more genes, remains largely unexplored, with only a very limited number of multi-gene SL sets validated experimentally to date. Because existing known SL interactions predominantly involve non-metabolic regulatory genes, they cannot be directly used to validate the metabolic higher-order SL sets identified in this study. This gap reflects a current limitation of the field rather than of our approach, and it underscores the need for future experimental work specifically targeting metabolic gene interactions.

Although our analyses provide systematic predictions of strictly selective metabolic vulnerabilities, several methodological limitations must be acknowledged. The findings rely entirely on genome-scale metabolic models and thus inherit assumptions such as steady-state flux balance and the absence of kinetic and regulatory data (and consequently the associated constraints). Context-specific reconstructions based on transcriptomic data may also introduce inaccuracies, and predictions of higher-order synthetic lethality remain sensitive to model structure and incomplete metabolic knowledge. These limitations emphasize that the predicted SL sets should be viewed as hypotheses requiring experimental validation. Future confirmation of the predicted SSDTs could be achieved using established combinatorial perturbation platforms, including CRISPR/Cas9 gene knockout screens, multiplex CRISPR systems enabling two- to four-gene targeting [91,92], and RNAi-based knockdown approaches such as siRNA or shRNA [93–95]. These perturbation tools can be applied in parallel to matched cancer and non-tumor control cell lines, followed by viability, apoptosis, or clonogenic assays to evaluate both lethality and cancer selectivity. Such validation workflows are commonly used for synthetic lethality testing and represent a practical next step for experimentally confirming the SSDTs, particularly the pan-cancer candidates.

Conclusion

In this work, for the first time, the superior capability of higher-order SL sets over single essential genes in providing strictly-selective drug targets was demonstrated using genome-scale metabolic models. Besides reporting over 500 solutions identified as strictly-selective for 13 types of cancers, nine quadruple SLs are specifically investigated as common solutions between five cancers. Targeting any of these nine SLs results in the disruption of glycolysis and the pentose phosphate pathway along with one of the sphingolipid metabolism, solute carrier family, amino-sugar metabolism, keratan sulfate biosynthesis, or nucleotide metabolism pathways. This simultaneous attack suggests a key to effectively and selectively targeting cancer cells. The provided results are solely based on in-silico experimentation, and further experimental studies are required in future work to examine the effectiveness of the obtained results.

Supporting information

S1 File. An Excel file with 14 sheets of data.

“Info” sheet describes the studied cancers by their TCGA name. The other sheets contain the identified strictly-selective genes and gene sets.

https://doi.org/10.1371/journal.pone.0342199.s001

(XLSX)

S2 File. Detailed information on all genes of potential SSDTs of the colon cancer.

https://doi.org/10.1371/journal.pone.0342199.s002

(PDF)

S3 File. The interaction graphs for LIHC and KIRC strictly-selective targets.

https://doi.org/10.1371/journal.pone.0342199.s003

(PDF)

References

  1. 1. Warburg O. Über den Stoffwechsel der Carcinomzelle. Naturwissenschaften. 1924;12(50):1131–7.
  2. 2. Cairns RA, Harris IS, Mak TW. Regulation of cancer cell metabolism. Nat Rev Cancer. 2011;11(2):85–95. pmid:21258394
  3. 3. Lunt SY, Vander Heiden MG. Aerobic glycolysis: meeting the metabolic requirements of cell proliferation. Annual Review of Cell and Developmental Biology. 2011;27:441–64.
  4. 4. Newsholme EA, Crabtree B, Ardawi MS. The role of high rates of glycolysis and glutamine utilization in rapidly dividing cells. Biosci Rep. 1985;5(5):393–400. pmid:3896338
  5. 5. Li Z, Zhang H. Reprogramming of glucose, fatty acid and amino acid metabolism for cancer progression. Cell Mol Life Sci. 2016;73(2):377–92. pmid:26499846
  6. 6. Costello LC, Franklin RB. “Why do tumour cells glycolyse?”: from glycolysis through citrate to lipogenesis. Mol Cell Biochem. 2005;280(1–2):1–8. pmid:16511951
  7. 7. Fernandez-de-Cossio-Diaz J, Vazquez A. Limits of aerobic metabolism in cancer cells. Sci Rep. 2017;7(1):13488. pmid:29044214
  8. 8. Fadaka A, Ajiboye B, Ojo O, Adewale O, Olayide I, Emuowhochere R. Biology of glucose metabolization in cancer cells. Journal of Oncological Sciences. 2017;3(2):45–51.
  9. 9. Gatenby RA, Gillies RJ. Glycolysis in cancer: a potential target for therapy. Int J Biochem Cell Biol. 2007;39(7–8):1358–66. pmid:17499003
  10. 10. Akram M. Mini-review on glycolysis and cancer. J Cancer Educ. 2013;28(3):454–7. pmid:23728993
  11. 11. Ganapathy-Kanniappan S, Geschwind J-FH. Tumor glycolysis as a target for cancer therapy: progress and prospects. Mol Cancer. 2013;12:152. pmid:24298908
  12. 12. Li X-B, Gu J-D, Zhou Q-H. Review of aerobic glycolysis and its key enzymes - new targets for lung cancer therapy. Thorac Cancer. 2015;6(1):17–24. pmid:26273330
  13. 13. Ganapathy-Kanniappan S. Molecular intricacies of aerobic glycolysis in cancer: current insights into the classic metabolic phenotype. Crit Rev Biochem Mol Biol. 2018;53(6):667–82. pmid:30668176
  14. 14. Hsu PP, Sabatini DM. Cancer cell metabolism: Warburg and beyond. Cell. 2008;134(5):703–7. pmid:18775299
  15. 15. DeBerardinis RJ, Mancuso A, Daikhin E, Nissim I, Yudkoff M, Wehrli S, et al. Beyond aerobic glycolysis: transformed cells can engage in glutamine metabolism that exceeds the requirement for protein and nucleotide synthesis. Proc Natl Acad Sci U S A. 2007;104(49):19345–50. pmid:18032601
  16. 16. Yang L, Venneti S, Nagrath D. Glutaminolysis: A Hallmark of Cancer Metabolism. Annu Rev Biomed Eng. 2017;19:163–94. pmid:28301735
  17. 17. Santos CR, Schulze A. Lipid metabolism in cancer. FEBS J. 2012;279(15):2610–23. pmid:22621751
  18. 18. Sun L, Suo C, Li S-T, Zhang H, Gao P. Metabolic reprogramming for cancer cells and their microenvironment: Beyond the Warburg Effect. Biochim Biophys Acta Rev Cancer. 2018;1870(1):51–66. pmid:29959989
  19. 19. TeSlaa T, Ralser M, Fan J, Rabinowitz JD. The pentose phosphate pathway in health and disease. Nat Metab. 2023;5(8):1275–89. pmid:37612403
  20. 20. Lunt SY, Fendt S-M. Metabolism – A cornerstone of cancer initiation, progression, immune evasion and treatment response. Current Opinion in Systems Biology. 2018;8:67–72.
  21. 21. Sharma A, Boise LH, Shanmugam M. Cancer Metabolism and the Evasion of Apoptotic Cell Death. Cancers (Basel). 2019;11(8):1144. pmid:31405035
  22. 22. Vander Heiden MG. Targeting cancer metabolism: a therapeutic window opens. Nature Reviews Drug Discovery. 2011;10(9):671–84.
  23. 23. Vernieri C, Casola S, Foiani M, Pietrantonio F, de Braud F, Longo V. Targeting Cancer Metabolism: Dietary and Pharmacologic Interventions. Cancer Discov. 2016;6(12):1315–33. pmid:27872127
  24. 24. Galluzzi L, Kepp O, Vander Heiden MG, Kroemer G. Metabolic targets for cancer therapy. Nat Rev Drug Discov. 2013;12(11):829–46. pmid:24113830
  25. 25. Nilsson A, Nielsen J. Genome scale metabolic modeling of cancer. Metab Eng. 2017;43(Pt B):103–12. pmid:27825806
  26. 26. Geng J, Nielsen J. In silico analysis of human metabolism: Reconstruction, contextualization and application of genome-scale models. Current Opinion in Systems Biology. 2017;2:29–38.
  27. 27. O’Brien EJ, Monk JM, Palsson BO. Using Genome-scale Models to Predict Biological Capabilities. Cell. 2015;161(5):971–87. pmid:26000478
  28. 28. Robinson JL, Nielsen J. Anticancer drug discovery through genome-scale metabolic modeling. Current Opinion in Systems Biology. 2017;4:1–8.
  29. 29. Ji Z, Yan K, Li W, Hu H, Zhu X. Mathematical and computational modeling in complex biological systems. BioMed research international. 2017;2017.
  30. 30. Folger O, Jerby L, Frezza C, Gottlieb E, Ruppin E, Shlomi T. Predicting selective drug targets in cancer through metabolic networks. Mol Syst Biol. 2011;7:501. pmid:21694718
  31. 31. Cruz F, Faria JP, Rocha M, Rocha I, Dias O. A review of methods for the reconstruction and analysis of integrated genome-scale models of metabolism and regulation. Biochem Soc Trans. 2020;48(5):1889–903. pmid:32940659
  32. 32. Yizhak K, Gaude E, Le Dévédec S, Waldman YY, Stein GY, van de Water B, et al. Phenotype-based cell-specific metabolic modeling reveals metabolic liabilities of cancer. Elife. 2014;3:e03641. pmid:25415239
  33. 33. Gatto F, Miess H, Schulze A, Nielsen J. Flux balance analysis predicts essential genes in clear cell renal cell carcinoma metabolism. Sci Rep. 2015;5:10738. pmid:26040780
  34. 34. Agren R, Bordel S, Mardinoglu A, Pornputtapong N, Nookaew I, Nielsen J. Reconstruction of genome-scale active metabolic networks for 69 human cell types and 16 cancer types using INIT. PLoS Comput Biol. 2012;8(5):e1002518. pmid:22615553
  35. 35. Barrena N, Valcárcel LV, Olaverri-Mendizabal D, Apaolaza I, Planes FJ. Synthetic lethality in large-scale integrated metabolic and regulatory network models of human cells. NPJ Syst Biol Appl. 2023;9(1):32. pmid:37454223
  36. 36. Larsson I, Uhlén M, Zhang C, Mardinoglu A. Genome-Scale Metabolic Modeling of Glioblastoma Reveals Promising Targets for Drug Development. Front Genet. 2020;11:381. pmid:32362913
  37. 37. Pacheco MP, Bintener T, Ternes D, Kulms D, Haan S, Letellier E, et al. Identifying and targeting cancer-specific metabolism with network-based drug target prediction. EBioMedicine. 2019;43:98–106. pmid:31126892
  38. 38. Zugazagoitia J, Guedes C, Ponce S, Ferrer I, Molina-Pinelo S, Paz-Ares L. Current Challenges in Cancer Treatment. Clin Ther. 2016;38(7):1551–66. pmid:27158009
  39. 39. O’Neil NJ, Bailey ML, Hieter P. Synthetic lethality and cancer. Nat Rev Genet. 2017;18(10):613–23. pmid:28649135
  40. 40. Huang A, Garraway LA, Ashworth A, Weber B. Synthetic lethality as an engine for cancer drug target discovery. Nat Rev Drug Discov. 2020;19(1):23–38. pmid:31712683
  41. 41. Yar MS, Haider K, Gohel V, Siddiqui NA, Kamal A. Synthetic lethality on drug discovery: an update on cancer therapy. Expert Opin Drug Discov. 2020;15(7):823–32. pmid:32228106
  42. 42. Kaelin WG Jr. The concept of synthetic lethality in the context of anticancer therapy. Nat Rev Cancer. 2005;5(9):689–98. pmid:16110319
  43. 43. Ngoi NYL, Gallo D, Torrado C, Nardo M, Durocher D, Yap TA. Synthetic lethal strategies for the development of cancer therapeutics. Nat Rev Clin Oncol. 2025;22(1):46–64. pmid:39627502
  44. 44. Frezza C, Zheng L, Folger O, Rajagopalan KN, MacKenzie ED, Jerby L, et al. Haem oxygenase is synthetically lethal with the tumour suppressor fumarate hydratase. Nature. 2011;477(7363):225–8. pmid:21849978
  45. 45. Zhang C, Ji B, Mardinoglu A, Nielsen J, Hua Q. Logical transformation of genome-scale metabolic models for gene level applications and analysis. Bioinformatics. 2015;31(14):2324–31. pmid:25735769
  46. 46. Agren R, Mardinoglu A, Asplund A, Kampf C, Uhlen M, Nielsen J. Identification of anticancer drugs for hepatocellular carcinoma through personalized genome-scale metabolic modeling. Mol Syst Biol. 2014;10(3):721. pmid:24646661
  47. 47. Feng Y, Long Y, Wang H, Ouyang Y, Li Q, Wu M, et al. Benchmarking machine learning methods for synthetic lethality prediction in cancer. Nat Commun. 2024;15(1):9058. pmid:39428397
  48. 48. Ahlmann-Eltze C, Huber W, Anders S. Deep-learning-based gene perturbation effect prediction does not yet outperform simple linear baselines. Nat Methods. 2025;22(8):1657–61. pmid:40759747
  49. 49. Swainston N, Smallbone K, Hefzi H, Dobson PD, Brewer J, Hanscho M. Recon 2.2: from reconstruction to model of human metabolism. Metabolomics. 2016;12:1–7.
  50. 50. Rahman M, Jackson LK, Johnson WE, Li DY, Bild AH, Piccolo SR. Alternative preprocessing of RNA-Sequencing data in The Cancer Genome Atlas leads to improved analysis results. Bioinformatics. 2015;31(22):3666–72. pmid:26209429
  51. 51. Opdam S, Richelle A, Kellman B, Li S, Zielinski DC, Lewis NE. A Systematic Evaluation of Methods for Tailoring Genome-Scale Metabolic Models. Cell Syst. 2017;4(3):318-329.e6. pmid:28215528
  52. 52. Machado D, Herrgård M. Systematic evaluation of methods for integration of transcriptomic data into constraint-based models of metabolism. PLoS Comput Biol. 2014;10(4):e1003580. pmid:24762745
  53. 53. Dehghan Manshadi M, Setoodeh P, Zare H. Rapid-SL identifies synthetic lethal sets with an arbitrary cardinality. Sci Rep. 2022;12(1):14022. pmid:35982201
  54. 54. Apaolaza I, Valcarcel LV, Planes FJ. gMCS: fast computation of genetic minimal cut sets in large networks. Bioinformatics. 2019;35(3):535–7. pmid:30052768
  55. 55. Pratapa A, Balachandran S, Raman K. Fast-SL: an efficient algorithm to identify synthetic lethal sets in metabolic networks. Bioinformatics. 2015;31(20):3299–305. pmid:26085504
  56. 56. Wishart DS, Feunang YD, Guo AC, Lo EJ, Marcu A, Grant JR, et al. DrugBank 5.0: a major update to the DrugBank database for 2018. Nucleic Acids Res. 2018;46(D1):D1074–82. pmid:29126136
  57. 57. Tang Z, Li C, Kang B, Gao G, Li C, Zhang Z. GEPIA: a web server for cancer and normal gene expression profiling and interactive analyses. Nucleic Acids Res. 2017;45(W1):W98–102. pmid:28407145
  58. 58. Groza T, Gomez FL, Mashhadi HH, Muñoz-Fuentes V, Gunes O, Wilson R, et al. The International Mouse Phenotyping Consortium: comprehensive knockout phenotyping underpinning the study of human disease. Nucleic Acids Res. 2023;51(D1):D1038–45. pmid:36305825
  59. 59. De Preter G, Neveu M-A, Danhier P, Brisson L, Payen VL, Porporato PE, et al. Inhibition of the pentose phosphate pathway by dichloroacetate unravels a missing link between aerobic glycolysis and cancer cell proliferation. Oncotarget. 2016;7(3):2910–20. pmid:26543237
  60. 60. Patra KC, Hay N. The pentose phosphate pathway and cancer. Trends Biochem Sci. 2014;39(8):347–54. pmid:25037503
  61. 61. Tian R-F, Li X-F, Xu C, Wu H, Liu L, Wang L-H, et al. SiRNA targeting PFK1 inhibits proliferation and migration and enhances radiosensitivity by suppressing glycolysis in colorectal cancer. Am J Transl Res. 2020;12(9):4923–40. pmid:33042398
  62. 62. Feng Y, Zhang Y, Cai Y, Liu R, Lu M, Li T, et al. A20 targets PFKL and glycolysis to inhibit the progression of hepatocellular carcinoma. Cell Death Dis. 2020;11(2):89. pmid:32015333
  63. 63. Li K, Zhu X, Yuan C. Inhibition of miR-185-3p Confers Erlotinib Resistance Through Upregulation of PFKL/MET in Lung Cancers. Front Cell Dev Biol. 2021;9:677860. pmid:34368128
  64. 64. Perl A, Hanczko R, Telarico T, Oaks Z, Landas S. Oxidative stress, inflammation and carcinogenesis are controlled through the pentose phosphate pathway by transaldolase. Trends Mol Med. 2011;17(7):395–403. pmid:21376665
  65. 65. Silvers CR, Miyamoto H, Messing EM, Netto GJ, Lee Y-F. Characterization of urinary extracellular vesicle proteins in muscle-invasive bladder cancer. Oncotarget. 2017;8(53):91199–208. pmid:29207636
  66. 66. Heinrich PC, Morris HP, Weber G. Behavior of transaldolase (EC 2.2.1.2) and transketolase (EC 2.2.1.1) Activities in normal, neoplastic, differentiating, and regenerating liver. Cancer Res. 1976;36(9 pt.1):3189–97. pmid:10080
  67. 67. Wang C, Guo K, Gao D, Kang X, Jiang K, Li Y, et al. Identification of transaldolase as a novel serum biomarker for hepatocellular carcinoma metastasis using xenografted mouse model and clinic samples. Cancer Lett. 2011;313(2):154–66. pmid:22023829
  68. 68. Risha Y, Susevski V, Hüttmann N, Poolsup S, Minic Z, Berezovski MV. Breast Cancer-Derived Microvesicles Are the Source of Functional Metabolic Enzymes as Potential Targets for Cancer Therapy. Biomedicines. 2021;9(2):107. pmid:33499132
  69. 69. Ding Y, Gong C, Huang D, Chen R, Sui P, Lin KH, et al. Synthetic lethality between HER2 and transaldolase in intrinsically resistant HER2-positive breast cancers. Nat Commun. 2018;9(1):4274. pmid:30323337
  70. 70. Sarfraz I, Rasul A, Hussain G, Shah MA, Zahoor AF, Asrar M, et al. 6-Phosphogluconate dehydrogenase fuels multiple aspects of cancer cells: From cancer initiation to metastasis and chemoresistance. Biofactors. 2020;46(4):550–62. pmid:32039535
  71. 71. Polat IH, Tarrado-Castellarnau M, Bharat R, Perarnau J, Benito A, Cortés R, et al. Oxidative Pentose Phosphate Pathway Enzyme 6-Phosphogluconate Dehydrogenase Plays a Key Role in Breast Cancer Metabolism. Biology (Basel). 2021;10(2):85. pmid:33498665
  72. 72. Yang X, Peng X, Huang J. Inhibiting 6-phosphogluconate dehydrogenase selectively targets breast cancer through AMPK activation. Clin Transl Oncol. 2018;20(9):1145–52. pmid:29340974
  73. 73. Zheng W, Feng Q, Liu J, Guo Y, Gao L, Li R, et al. Inhibition of 6-phosphogluconate Dehydrogenase Reverses Cisplatin Resistance in Ovarian and Lung Cancer. Front Pharmacol. 2017;8:421. pmid:28713273
  74. 74. Khan GB, Qasim M, Rasul A, Ashfaq UA, Alnuqaydan AM. Identification of Lignan Compounds as New 6-Phosphogluconate Dehydrogenase Inhibitors for Lung Cancer. Metabolites. 2022;13(1):34. pmid:36676959
  75. 75. Liu R, Li W, Tao B, Wang X, Yang Z, Zhang Y, et al. Tyrosine phosphorylation activates 6-phosphogluconate dehydrogenase and promotes tumor growth and radiation resistance. Nat Commun. 2019;10(1):991. pmid:30824700
  76. 76. Xu J, Ren G, Cheng Q. Inhibition of 6-Phosphogluconate Dehydrogenase Reverses Epirubicin Resistance Through Metabolic Reprograming in Triple-Negative Breast Cancer Cells. Technol Cancer Res Treat. 2023;22:15330338231190737. pmid:37559469
  77. 77. Ma L, Cheng Q. Inhibiting 6-phosphogluconate dehydrogenase reverses doxorubicin resistance in anaplastic thyroid cancer via inhibiting NADPH-dependent metabolic reprogramming. Biochem Biophys Res Commun. 2018;498(4):912–7. pmid:29534964
  78. 78. Machala M, Procházková J, Hofmanová J, Králiková L, Slavík J, Tylichová Z, et al. Colon Cancer and Perturbations of the Sphingolipid Metabolism. Int J Mol Sci. 2019;20(23):6051. pmid:31801289
  79. 79. Ogretmen B. Sphingolipid metabolism in cancer signalling and therapy. Nat Rev Cancer. 2018;18(1):33–50. pmid:29147025
  80. 80. Modrak DE, Gold DV, Goldenberg DM. Sphingolipid targets in cancer therapy. Mol Cancer Ther. 2006;5(2):200–8. pmid:16505092
  81. 81. Shaw J, Costa-Pinheiro P, Patterson L, Drews K, Spiegel S, Kester M. Novel Sphingolipid-Based Cancer Therapeutics in the Personalized Medicine Era. Adv Cancer Res. 2018;140:327–66. pmid:30060815
  82. 82. Truman J-P, García-Barros M, Obeid LM, Hannun YA. Evolving concepts in cancer therapy through targeting sphingolipid metabolism. Biochim Biophys Acta. 2014;1841(8):1174–88. pmid:24384461
  83. 83. Janneh AH, Ogretmen B. Targeting Sphingolipid Metabolism as a Therapeutic Strategy in Cancer Treatment. Cancers (Basel). 2022;14(9):2183. pmid:35565311
  84. 84. Li Q, Shu Y. Role of solute carriers in response to anticancer drugs. Mol Cell Ther. 2014;2:15. pmid:26056583
  85. 85. Cheng H, Wang S, Gao D, Yu K, Chen H, Huang Y, et al. Nucleotide sugar transporter SLC35A2 is involved in promoting hepatocellular carcinoma metastasis by regulating cellular glycosylation. Cell Oncol (Dordr). 2023;46(2):283–97. pmid:36454514
  86. 86. Xu S, Chen X, Fang J, Chu H, Fang S, Zeng L, et al. Comprehensive analysis of 33 human cancers reveals clinical implications and immunotherapeutic value of the solute carrier family 35 member A2. Front Immunol. 2023;14:1155182. pmid:37275857
  87. 87. O’Day EM, Idos GE, Hill C, Chen JW, Wagner G. Cytidine monophosphate N-acetylneuraminic acid synthetase enhances invasion of human triple-negative breast cancer cells. Onco Targets Ther. 2018;11:6827–38. pmid:30349315
  88. 88. Ho M-L, Kuo W-K, Chu LJ, Yeh I-H, Fan C-W, Chang H-I, et al. N-acetylgalactosamine-6-sulfatase (GALNS), Similar to Glycodelin, Is a Potential General Biomarker for Multiple Malignancies. Anticancer Res. 2019;39(11):6317–24. pmid:31704862
  89. 89. Zhang Z, Wu H, Chen Z, Li G, Liu B. Circular RNA ATXN7 promotes the development of gastric cancer through sponging miR-4319 and regulating ENTPD4. Cancer Cell Int. 2020;20:25. pmid:31997941
  90. 90. Previtali V, Bagnolini G, Ciamarone A, Ferrandi G, Rinaldi F, Myers SH, et al. New Horizons of Synthetic Lethality in Cancer: Current Development and Future Perspectives. J Med Chem. 2024;67(14):11488–521. pmid:38955347
  91. 91. Shalem O, Sanjana NE, Hartenian E, Shi X, Scott DA, Mikkelson T, et al. Genome-scale CRISPR-Cas9 knockout screening in human cells. Science. 2014;343(6166):84–7. pmid:24336571
  92. 92. Najm FJ, Strand C, Donovan KF, Hegde M, Sanson KR, Vaimberg EW. Orthologous CRISPR–Cas9 enzymes for combinatorial genetic screens. Nature Methods. 2018;15(8):671–4.
  93. 93. Kampmann M, Bassik MC, Weissman JS. Design and analysis of genome-wide pooled shRNA screens. Cell Reports. 2014;9(5):2079–90.
  94. 94. Jackson AL, Linsley PS. Recognizing and avoiding siRNA off-target effects for target identification and therapeutic application. Nat Rev Drug Discov. 2010;9(1):57–67. pmid:20043028
  95. 95. Mohr S, Bakal C, Perrimon N. Genomic screening with RNAi: results and challenges. Nature Reviews Molecular Cell Biology. 2014;15(11):674–84.