Transfer learning for drug-target interaction prediction.

MOTIVATION: Utilizing AI-driven approaches for drug-target interaction (DTI) prediction require large volumes of training data which are not available for the majority of target proteins. In this study, we investigate the use of deep transfer learning for the prediction of interactions between drug candidate compounds and understudied target proteins with scarce training data. The idea here is to first train a deep neural network classifier with a generalized source training dataset of large size and then to reuse this pre-trained neural network as an initial configuration for re-training/fine-tuning purposes with a small-sized specialized target training dataset. To explore this idea, we selected six protein families that have critical importance in biomedicine: kinases, G-protein-coupled receptors (GPCRs), ion channels, nuclear receptors, proteases, and transporters. In two independent experiments, the protein families of transporters and nuclear receptors were individually set as the target datasets, while the remaining five families were used as the source datasets. Several size-based target family training datasets were formed in a controlled manner to assess the benefit provided by the transfer learning approach. RESULTS: Here, we present a systematic evaluation of our approach by pre-training a feed-forward neural network with source training datasets and applying different modes of transfer learning from the pre-trained source network to a target dataset. The performance of deep transfer learning is evaluated and compared with that of training the same deep neural network from scratch. We found that when the training dataset contains fewer than 100 compounds, transfer learning outperforms the conventional strategy of training the system from scratch, suggesting that transfer learning is advantageous for predicting binders to under-studied targets. AVAILABILITY AND IMPLEMENTATION: The source code and datasets are available at https://github.com/cansyl/TransferLearning4DTI. Our web-based service containing the ready-to-use pre-trained models is accessible at https://tl4dti.kansil.org.

Context dependent isoform specific PI3K inhibition confers drug resistance in hepatocellular carcinoma cells.

BACKGROUND: Targeted therapies for Primary liver cancer (HCC) is limited to the multi-kinase inhibitors, and not fully effective due to the resistance to these agents because of the heterogeneous molecular nature of HCC developed during chronic liver disease stages and cirrhosis. Although combinatorial therapy can increase the efficiency of targeted therapies through synergistic activities, isoform specific effects of the inhibitors are usually ignored. This study concentrated on PI3K/Akt/mTOR pathway and the differential combinatory bioactivities of isoform specific PI3K-α inhibitor (PIK-75) or PI3K-ß inhibitor (TGX-221) with Sorafenib dependent on PTEN context. METHODS: The bioactivities of inhibitors on PTEN adequate Huh7 and deficient Mahlavu cells were investigated with real time cell growth, cell cycle and cell migration assays. Differentially expressed genes from RNA-Seq were identified by edgeR tool. Systems level network analysis of treatment specific pathways were performed with Prize Collecting Steiner Tree (PCST) on human interactome and enriched networks were visualized with Cytoscape platform. RESULTS: Our data from combinatory treatment of Sorafenib and PIK-75 and TGX-221 showed opposite effects; while PIK-75 displays synergistic effects on Huh7 cells leading to apoptotic cell death, Sorafenib with TGX-221 display antagonistic effects and significantly promotes cell growth in PTEN deficient Mahlavu cells. Signaling pathways were reconstructed and analyzed in-depth from RNA-Seq data to understand mechanism of differential synergistic or antagonistic effects of PI3K-α (PIK-75) and PI3K-ß (TGX-221) inhibitors with Sorafenib. PCST allowed as to identify AOX1 and AGER as targets in PI3K/Akt/mTOR pathway for this combinatory effect. The siRNA knockdown of AOX1 and AGER significantly reduced cell proliferation in HCC cells. CONCLUSIONS: Simultaneously constructed and analyzed differentially expressed cellular networks presented in this study, revealed distinct consequences of isoform specific PI3K inhibition in PTEN adequate and deficient liver cancer cells. We demonstrated the importance of context dependent and isoform specific PI3K/Akt/mTOR signaling inhibition in drug resistance during combination therapies. ( https://github.com/cansyl/Isoform-spesific-PI3K-inhibitor-analysis ).

Design, Synthesis and Biological Evaluation of Novel Triazolothiadiazoles Derived from NSAIDs as Anticancer Agents.

BACKGROUND: Although transplantation, surgical resection, and tumor ablation are treatment options available following early diagnosis of HCC, poor prognosis and high recurrence rates restrict the efficacy of these approaches. Hence, small molecules with high selectivity and bioactivity are urgently required. OBJECTIVE: This study presents the synthesis of a series of new triazolothiadiazole derivatives (1a-3j) with NSAID moieties and their cytotoxic bioactivities. METHODS: The new synthetic derivatives (1-3; 1a-3j) and NSAIDs ibuprofen, naproxen, and flurbiprofen that commonly used in clinics were screened against human liver (Huh7), breast (MCF7), and colon (HCT116) carcinoma cell lines under in vitro conditions via NCI-sulforhodamine B assay. RESULTS: The 4-methoxyphenyl substituted condensed derivatives 1h, 2h, and 3h were the most active compounds. Based on its high potency, compound 3h was selected for the further biological evaluation of hepatocellular carcinoma cell lines, and the mechanisms underlying cell death induced by 3h were determined. The results revealed that compound 3h induced apoptosis and cell cycle arrest in the sub G1 phase in human liver cancer cells. CONCLUSION: These new small molecules may be used for the development of new lead compounds.

Design, Synthesis and In Vitro Cytotoxic Activity of New 6,9-Disubstituted Purine Analogues.

A series of new 6,9-disubstituted purine analogs with 4-substituted piperazine at C-6 and 4-substituted benzyl at N-9 were designed and synthesized in four steps. All synthesized compounds (7-26) were screened initially for their in vitro anticancer activity on Huh7 liver, HCT116 colon and MCF7 breast carcinoma cell lines. Cytotoxic bioactivity studies revealed that all compounds screened, with compound 19 being the exception, were found to have promising cytotoxic activities at IC50 range of 0.05-21.8 ∆M against cancer cells Huh7, HCT116 and MCF7. Among the prepared purine analogs, two of them (12 and 22) exhibited excellent cytotoxic activities, with IC50 0.08-0.13 ∆M, on Huh7 cells comparable to camptothecin (CPT) and better than cladribine, fludarabine and 5-FU. Afterwards, the evaluation of cytotoxicity of the most potent purine analogs was screened against further hepatocellular cancer (HCC) cell lines. The 6-(4-(4-trifluoromethylphenyl)piperazine (12) and 6-(4-(3,4-dichlorophenyl)piperazine analogs (25) displayed a significant IC50 values (IC50 < 0.1-0.13 ∆M) comparable to CPT and better cytotoxic bioactivity when compared with 5-FU, cladribine and fludarabine on HCC cells (Huh7 and HepG2).

Development of a novel zebrafish xenograft model in ache mutants using liver cancer cell lines.

Acetylcholinesterase (AChE), an enzyme responsible for degradation of acetylcholine, has been identified as a prognostic marker in liver cancer. Although in vivo Ache tumorigenicity assays in mouse are present, no established liver cancer xenograft model in zebrafish using an ache mutant background exists. Herein, we developed an embryonic zebrafish xenograft model using epithelial (Hep3B) and mesenchymal (SKHep1) liver cancer cell lines in wild-type and ache sb55 sibling mutant larvae after characterization of cholinesterase expression and activity in cell lines and zebrafish larvae. The comparison of fluorescent signal reflecting tumor size at 3-days post-injection (dpi) revealed an enhanced tumorigenic potential and a reduced migration capacity in cancer cells injected into homozygous ache sb55 mutants when compared with the wild-type. Increased tumor load was confirmed using an ALU based tumor DNA quantification method modified for use in genotyped xenotransplanted zebrafish embryos. Confocal microscopy using the Huh7 cells stably expressing GFP helped identify the distribution of tumor cells in larvae. Our results imply that acetylcholine accumulation in the microenvironment directly or indirectly supports tumor growth in liver cancer. Use of this model system for drug screening studies holds potential in discovering new cholinergic targets for treatment of liver cancers.

Synthesis of Novel Benzothiazole-Piperazine Derivatives and Their Biological Evaluation as Acetylcholinesterase Inhibitors and Cytotoxic Agents.

OBJECTIVE AND METHOD: A new series of benzothiazole-piperazine derivatives was synthesized and a complete chemical characterization of the novel compounds was provided. In vitro cytotoxic activities were screened against colorectal (HCT-116), breast (MCF-7) and hepatocellular (Huh7) cancer cell lines by Sulforhodamine B assay. RESULT AND DISCUSSION: All compounds showed cytotoxic activity against hepatocellular (Huh7) and breast (MCF-7) cancer cell lines. Dihalo substituted benzylpiperazine derivatives (2a, 2e) had the highest cytotoxic activities in all the tested cell lines. In addition, acetylcholinesterase (AChE) and butyrylcholinesterase (BuChE) inhibitory activities of synthesized compounds were investigated by in vitro Ellman's method. Compound 2j led to moderate and selective inhibition against AChE. Docking study was utilized to understand the binding mode of compound 2j in comparision with donepezil on AChE. The other tested compounds showed weak or no inhibition against AChE as promising anticancer agents.

Synthesis of novel substituted purine derivatives and identification of the cell death mechanism.

Novel 9-(substituted amino/piperazinoethyl)adenines (4-12), 6-(substituted piperazino/amino)purines (15-27), 9-(p-toluenesulfonyl/cyclopentyl/ethoxycarbonylmethyl)-6-(substituted amino/piperazino)purines (28-34, 36, 37, 38-41) were synthesized and evaluated initially for their cytotoxic activities on liver Huh7, breast T47D and colon HCT116 carcinoma cells. N(6)-(4-Trifluoromethylphenyl)piperazine derivative (17) and its 9-(p-toluene-sulfonyl)/9-cyclopentyl analogues (28, 36) had promising cytotoxic activities. Compounds 17, 28 and 36 were further analysed for their cytotoxicity in a panel of a liver cancer cell lines. The compound 36 had better cytotoxic activities (IC50 ≤ 1 µM) than the nucleobase 5-FU and nucleosides fludarabine, cladribine, and pentostatine on Huh7 cells. Cytotoxicity induced by 36 was later identified as senescence associated cell death by SA-ß-Gal assay.

Bi-k-bi clustering: mining large scale gene expression data using two-level biclustering.

Due to the increase in gene expression data sets in recent years, various data mining techniques have been proposed for mining gene expression profiles. However, most of these methods target single gene expression data sets and cannot handle all the available gene expression data in public databases in reasonable amount of time and space. In this paper, we propose a novel framework, bi-k-bi clustering, for finding association rules of gene pairs that can easily operate on large scale and multiple heterogeneous data sets. We applied our proposed framework on the available NCBI GEO Homo sapiens data sets. Our results show consistency and relatedness with the available literature and also provides novel associations.

A novel form of recessive limb girdle muscular dystrophy with mental retardation and abnormal expression of alpha-dystroglycan.

The limb girdle muscular dystrophies are a heterogeneous group of conditions characterized by proximal muscle weakness and disease onset ranging from infancy to adulthood. We report here eight patients from seven unrelated families affected by a novel and relatively mild form of autosomal recessive limb girdle muscular dystrophy (LGMD2) with onset in the first decade of life and characterized by severe mental retardation but normal brain imaging. Immunocytochemical studies revealed a significant selective reduction of alpha-dystroglycan expression in the muscle biopsies. Linkage analysis excluded known loci for both limb girdle muscular dystrophy and congenital muscular dystrophies in the consanguineous families. We consider that this represents a novel form of muscular dystrophy with associated brain involvement. The biochemical studies suggest that it may belong to the growing number of muscular dystrophies with abnormal expression of alpha-dystroglycan.