Big Data to Knowledge: Application of Machine Learning to Predictive Modeling of Therapeutic Response in Cancer

BackgroundIn recent years, the availability of high throughput technologies, establishment of large molecular patient data repositories, and advancement in computing power and storage have allowed elucidation of complex mechanisms implicated in therapeutic response in cancer patients. The breadth and depth of such data, alongside experimental noise and missing values, requires a sophisticated human-machine interaction that would allow effective learning from complex data and accurate forecasting of future outcomes, ideally embedded in the core of machine learning design.ObjectiveIn this review, we will discuss machine learning techniques utilized for modeling of treatment response in cancer, including Random Forests, support vector machines, neural networks, and linear and logistic regression. We will overview their mathematical foundations and discuss their limitations and alternative approaches in light of their application to therapeutic response modeling in cancer.ConclusionWe hypothesize that the increase in the number of patient profiles and potential temporal monitoring of patient data will define even more complex techniques, such as deep learning and causal analysis, as central players in therapeutic response modeling.     

Molecular dynamics insights for PI3K-δ inhibition & structure guided identification of novel PI3K-δ inhibitors

Conjugated structure based and ligand based drug design techinques have been used previously to unearth putative binding ligands for kinase inhibition. PI3K-δ is a lipid kinase and it has been found abberant in diseases such as cancer,inflammation etc. Preliminarily, protein crystal structure analysis suggest avaibility of two crystal structures with varying degree of root mean square de throughtion in protein back bone and root mean square fluctuation in side chain geometry. Therefore, PI3K-δ crystal structure was selected based on charactristic reciever operating characterstic curve and % enrichment of actives analysis. Active site analysis through molecular dynamics simulations provided insights about four residues Ile910, Asp911, Met752, Lys755, which act as flap. These residues fecilitate ligand binding in a unique manner.Thereafter, a validated designed protocol has been used to screen asinex ligand database using molecular docking and binding energy calculations. Based on binding affinity & energy scores and interaction pattern analysis total top 50 ligands were selected for PI3K-δ inhibition studies. Moreover, two molecules ethyl 2-(2-((4-chloro-1-methyl-1H-pyrazole-3-carbonyl) oxy)acetamido) benzo[1]thiazole-6-carboxylate and 1,6,7-trimethyl-8-((tetrahydrofuran-2-yl) methyl)-1H-imidazo [1',2':1,5] pyrrolo[3,2-d]pyrimidine-2,4(3H,8H)-dione have been identified, which could be potential hits for PI3K-δ using insights provided by molecular modelling studies. The identified compunds were subjected to pan assay interference compound filter and were found to be compliant. Quantum mechanical calculations were perfromed for identified hits. The above strategy could be implemented as a strategy for rational drug design.

Communicated by Ramaswamy H. Sarma     

Oral tolerance induction with antigen conjugated to cholera toxin B subunit generates both Foxp3+CD25+ and Foxp3-CD25- CD4+ regulatory T cells

Oral administration of Ag coupled to cholera toxin B subunit (CTB) efficiently induces peripheral immunological tolerance. We investigated the extent to which this oral tolerance is mediated by CD25+CD4+ regulatory T cells (T(reg)). We found that total T(reg), KJ1-26+ T(reg) and CTLA-4+ T(reg) were all increased in Peyer's patches, mesenteric lymph nodes, and, to a lesser extent, in spleen of mice after intragastric administration of OVA/CTB conjugate, which also increased TGF-beta in serum. This could be abolished by co-administering cholera toxin or by treatment with anti-TGF-beta mAb. CD25+ T(reg), but also CD25-CD4+ T cells from OVA/CTB-treated BALB/c or DO11.10 mice efficiently suppressed effector T cell proliferation and IL-2 production in vitro. Following adoptive transfer, both T cell populations also suppressed OVA-specific T cell and delayed-type hypersensitivity responses in vivo. Foxp3 was strongly expressed by CD25+ T(reg) from OVA/CTB-treated mice, and treatment also markedly expanded CD25+Foxp3+ T(reg). Furthermore, in Rag1(-/-) mice that had adoptively received highly purified Foxp3-CD25-CD4+ OT-II T cells OVA/CTB feeding efficiently induced CD25+ T(reg) cells, which expressed Foxp3 more strongly than naturally developing T(reg) and also had stronger ability to suppress effector OT-II T cell proliferation. A remaining CD25- T cell population, which also became suppressive in response to OVA/CTB treatment, did not express Foxp3. Our results demonstrate that oral tolerance induced by CTB-conjugated Ag is associated with increase in TGF-beta and in both the frequency and suppressive capacity of Foxp3+ and CTLA-4+ CD25+ T(reg) together with the generation of both Foxp3+ and Foxp3-CD25- CD4+ T(reg).     

Comprehensive structural modeling and preparation of human 5-HT2A G-protein coupled receptor in functionally active form

The serotonin 2A receptor (5-HT_2AR) is an important member of the G-protein coupled receptor (GPCR) family involved in an array of neuromodulatory functions. Although the high-resolution structures of truncated versions of GPCRs, captured in ligand-bound conformational states, are available, the structures lack several functional regions, which have crucial roles in receptor response. Here, in order to understand the structure and dynamics of the ligand-free form of the receptor, we have performed meticulous modeling of the 5-HT_2AR with the third intracellular loop (ICL3). Our analyses revealed that the ligand-free ground state structure of 5-HT_2AR has marked distinction with ligand-bound conformations of 5-HT₂ subfamily proteins and exhibits extensive backbone flexibility across the loop regions, suggesting the importance of purifying the receptor in its native form for further studies. Hence, we have standardized a strategy that efficiently increases the expression of 5-HT_2AR by infecting Sf9 cells with a very low multiplicity of infection of baculovirus in conjunction with production boost additive and subsequently, purify the full-length receptor. Furthermore, we have optimized the selective over-expression of glycosylated and nonglycosylated forms of the receptor merely by switching the postinfection growth time, a method that has not been reported earlier.     

Cloning and Characterization of a Gene Responsible for the Lathyrus sativus Neurotoxin Degradation

A plasmid pBYA1, capable of degrading the Lathyrus sativus neurotoxin β-N-oxalyl amino alanine (BOAA), was isolated from a soli microbe and a physical map of the plasmid constructed. A partial Sau3A library of the plasmid was then prepared in pUC18 and a recombinant clone with a 1.8 kb insert capable of growing in M9 medium, with BOAA as sole source of carbon and nitrogen, was isolated. A nested set of deletions of this 1.8 kb fragment was then generated using Bal31 and the shortened fragments subcloned in the vector pUC18. Each of these clones were sequenced by Sanger’s dideoxy method and the sequences overlapped and analysed. The 1.8 kb BOAA degrading fragment was found to contain a 630 nucleotide long coding stretch, coding for 191 amino acids. The initial 56 nucleotides were found to contain sequences resembling the Pribnow-Schaller box of prokaryotes, the “-35” sequences and a sequence resembling the “-43” region of the lac promoter.     

Identification and characterization of Ixodes scapularis antigens that elicit tick immunity using yeast surface display

Repeated exposure of rabbits and other animals to ticks results in acquired resistance or immunity to subsequent tick bites and is partially elicited by antibodies directed against tick antigens. In this study we demonstrate the utility of a yeast surface display approach to identify tick salivary antigens that react with tick-immune serum. We constructed an Ixodes scapularis nymphal salivary gland yeast surface display library and screened the library with nymph-immune rabbit sera and identified five salivary antigens. Four of these proteins, designated P8, P19, P23 and P32, had a predicted signal sequence. We generated recombinant (r) P8, P19 and P23 in a Drosophila expression system for functional and immunization studies. rP8 showed anti-complement activity and rP23 demonstrated anti-coagulant activity. Ixodes scapularis feeding was significantly impaired when nymphs were fed on rabbits immunized with a cocktail of rP8, rP19 and rP23, a hall mark of tick-immunity. These studies also suggest that these antigens may serve as potential vaccine candidates to thwart tick feeding.     

Ixodes scapularis salivary gland protein P11 facilitates migration of Anaplasma phagocytophilum from the tick gut to salivary glands

Ixodes ticks harbour several human pathogens belonging to the order Rickettsiales, including Anaplasma phagocytophilum, the agent of human anaplasmosis. When ticks feed on A. phagocytophilum-infected mice, the pathogen enters the ticks' gut. The bacteria then migrate from the gut to infect the salivary glands of the ticks and are transmitted to the next host via the saliva. The molecular mechanisms that enable the migration of A. phagocytophilum from the gut to the salivary glands are poorly understood. Here we show that a secreted tick protein, P11, is important in this process. We show that P11 enables A. phagocytophilum to infect tick haemocytes, which are required for the migration of A. phagocytophilum from the gut to the salivary glands. Silencing of p11 impaired the A. phagocytophilum infection of tick haemocytes in vivo and consequently decreased pathogen infection of the salivary glands. In vitro experiments showed that P11 could bind to A. phagocytophilum and thus facilitate its infection of tick cells. This report provides new insights into A. phagocytophilum infection of ticks and reveals new avenues to interrupt the life cycle of Anaplasma and related Rickettsial pathogens.     

Normalization of deranged signal transduction in lymphocytes of COPD patients by the novel calcium channel blocker H-DHPM

Investigations on the role of intracellular Ca²⁺ ion concentration in the mechanism of development of COPD in smokers and non-smokers were carried out. The intracellular Ca²⁺ levels were found to be increased in human lymphocytes in patients with COPD as compared to non-smokers and smokers without COPD. The investigations reveal an association in altered intracellular Ca²⁺ regulation in lymphocytes and severity of COPD, by means of significant activation of Protein kinase C and inducible nitric oxide synthase (iNOS). The effect of a novel calcium channel blocker ethyl 4-(4′-heptanoyloxyphenyl)-6-methyl-3,4-dihydropyrimidin-2-one-5-carboxylate (H-DHPM) as a potential candidate for the treatment of COPD was also investigated. H-DHPM treated cells showed a decrease in intracellular Ca²⁺ level as compared to the control cells. Molecular studies were carried out to evaluate the expression profile of NOS isoforms in human lymphocytes and it was shown that H-DHPM decreases the increased iNOS in COPD along with reestablishing the normal levels of endothelial nitric oxide synthase (eNOS). The results of H-DHPM were comparable with those of Amlodipine, a known calcium channel blocker. Calcium channel blocker H-DHPM proves to be a potential candidate for the treatment of COPD and further clinical studies are required to prove its role in the treatment of pulmonary hypertension (PH).     

Identification and characterization of a novel legume-like lectin cDNA sequence from the red marine algae <Emphasis Type="Italic">Gracilaria fisheri</Emphasis>

A legume-type lectin (L-lectin) gene of the red algae Gracilaria fisheri (GFL) was cloned by rapid amplification of cDNA ends (RACE). The full-length cDNA of GFL was 1714 bp and contained a 1542 bp open reading frame encoding 513 amino acids with a predicted molecular mass of 56.5 kDa. Analysis of the putative amino acid sequence with NCBI-BLAST revealed a high homology (30–68%) with legume-type lectins (L-lectin) from Griffithsia japonica, Clavispora lusitaniae, Acyrthosiphon pisum, Tetraodon nigroviridis and Xenopus tropicalis. Phylogenetic relationship analysis showed the highest sequence identity to a glycoprotein of the red algae Griffithsia japonica (68%) (GenBank number AAM93989). Conserved Domain Database analysis detected an N-terminal carbohydrate recognition domain (CRD), the characteristic of L-lectins, which contained two sugar binding sites and a metal binding site. The secondary structure prediction of GFL showed a β-sheet structure, connected with turn and coil. The most abundant structural element of GFL was the random coil, while the α-helixes were distributed at the N- and C-termini, and 21 β-sheets were distributed in the CRD. Computer analysis of three-dimensional structure showed a common feature of L-lectins of GFL, which included an overall globular shape that was composed of a β-sandwich of two anti-parallel β-sheets, monosaccharide binding sites, were on the top of the structure and in proximity with a metal binding site. Northern blot analysis using a DIG-labelled probe derived from a partial GFL sequence revealed a hybridization signal of ~1.7 kb consistent with the length of the full-length GFL cDNA identified by RACE. No detectable band was observed from control total RNA extracted from filamentous green algae.