External Validation Series Research Articles

Artificial intelligence-based pathology as a biomarker of sensitivity to atezolizumab–bevacizumab in patients with hepatocellular carcinoma: a multicentre retrospective study

Clinical benefits of atezolizumab plus bevacizumab (atezolizumab-bevacizumab) are observed only in a subset of patients with hepatocellular carcinoma and the development of biomarkers is needed to improve therapeutic strategies. The atezolizumab-bevacizumab response signature (ABRS), assessed by molecular biology profiling techniques, has been shown to be associated with progression-free survival after treatment initiation. The primary objective of our study was to develop an artificial intelligence (AI) model able to estimate ABRS expression directly from histological slides, and to evaluate if model predictions were associated with progression-free survival. In this multicentre retrospective study, we developed a model (ABRS-prediction; ABRS-P), which was derived from the previously published clustering-constrained attention multiple instance learning (or CLAM) pipeline. We trained the model fit for regression analysis using a multicentre dataset from The Cancer Genome Atlas (patients treated by surgical resection, n=336). The ABRS-P model was externally validated on two independent series of samples from patients with hepatocellular carcinoma (a surgical resection series, n=225; and a biopsy series, n=157). The predictive value of the model was further tested in a series of biopsy samples from a multicentre cohort of patients with hepatocellular carcinoma treated with atezolizumab-bevacizumab (n=122). All samples in the study were from adults (aged ≥18 years). The validation sets were sampled between Jan 1, 2008, to Jan 1, 2023. For the multicentre validation set, the primary objective was to assess the association of high versus low ABRS-P values, defined relative to cross-validation median split thresholds in the first biopsy series, with progression-free survival after treatment initiation. Finally, we performed spatial transcriptomics and matched prediction heatmaps with in situ expression profiles. Of the 840 patients sampled, 641 (76%) were male and 199 (24%) were female. Across the development and validation datasets, hepatocellular carcinoma risk factors included alcohol intake, hepatitis B and C virus infections, and non-alcoholic steatohepatitis. Using cross-validation in the development series, the mean Pearson's correlation between ABRS-P values and ABRS score (mean expression of ABRS genes) was r=0·62 (SD 0·09; mean p<0·0001, SD<0·0001). The ABRS-P generalised well on the external validation series (surgical resection series, r=0·60 [95% CI 0·51-0·68], p<0·0001; biopsy series, r=0·53 [0·40-0·63], p<0·0001). In the 122 patients treated with atezolizumab-bevacizumab, those with ABRS-P-high tumours (n=74) showed significantly longer median progression-free survival than those with ABRS-P-low tumours (n=48) after treatment initiation (12 months [95% CI 7-not reached] vs 7 months [4-9]; p=0·014). Spatial transcriptomics showed significantly higher ABRS score, along with upregulation of various other immune effectors, in tumour areas with high ABRS-P values versus areas with low ABRS-P values. Our study indicates that AI applied on hepatocellular carcinoma digital slides is able to serve as a biomarker for progression-free survival in patients treated with atezolizumab-bevacizumab. This approach could be used in the development of inexpensive and fast biomarkers for targeted therapies. The combination of AI heatmaps with spatial transcriptomics provides insight on the molecular features associated with predictions. This methodology could be applied to other cancers or diseases and improve understanding of the biological mechanisms that drive responses to treatments. Institut National du Cancer, Fondation ARC, China Scholarship Council, Ligue Contre le Cancer du Val de Marne, Fondation de l'Avenir, Ipsen, and Fondation Bristol Myers Squibb Pour la Recherche en Immuno-Oncologie.

Validation of a post operative complication risk prediction algorithm in a non-head and neck squamous cell carcinoma cohort

Risk-adjusted algorithms in surgical audit attempt to adjust for patient case mix and complexity in order that published outcomes fairly reflect surgical performance and quality of care. Such risk-adjustment models have applied to head and neck squamous cell carcinoma (HNSCC). We test one algorithm, currently embedded in the oncology and reconstruction dataset within the Quality and Outcomes in Oral and Maxillofacial (QOMS) Audit, which is an artificial neural network, for its predictive accuracy on a surgical cohort receiving curative surgery for non-HNSCC pathology. A single centre retrospective case note audit of post operative complications between 2010 and 2020 was conducted on patients having curative surgery for non-HNSCC pathology. The observed complication rate was compared to the predicted probability of complications in order to test the performance of the complication risk-adjustment model. Of 1591 non-HNSCC patients, 58 met the inclusion criteria with a 30-day complication rate of 8/58 (13%). The artificial neural network predicted a complication rate of 20/58 (27%). Sensitivity (0.75), specificity (0.72) and overall accuracy (0.72) suggest acceptable discrimination. Hosmer-Lemershow Goodness of Fit test was good (p = 0.55) suggesting acceptable calibration though over-prediction of complication rate in the highest risk patents was observed. This external validation series suggests the algorithm can be applied to the non-HNSCC cohort, though some refinement of the algorithm is required to account for over-prediction of complications for higher-risk patients. With further analysis a robust means of risk adjusting for the non-HNSCC cohort should be possible.

Establishment and Validation of an Interferon-Stimulated Genes (ISGs) Prognostic Signature in Pan-cancer Patients: A Multicenter, Real-world Study.

Our study aims at developing an interferon-stimulated genes (ISGs) signature that could predict overall survival (OS) in cancer patients, which enrolled a total of 5643 pan-cancer patients. Linear models for microarray data method analysis were conducted to identify the differentially expressed prognostic genes in the global ISGs family. Time-dependent receiver operating characteristic (ROC) and Kaplan-Meier survival analysis were used to test the efficiency of a multi-gene signature in predicting the prognosis of pan-cancer patients. The prognostic performance and potential biological function of gene signature were verified by quantitative real-time PCR in a pan-cancer independent cohort. Three ISGs genes were finally identified to build a classifier, a specific risk score formula, with which patients were classified into the low- or high-risk groups. Time-dependent ROC analyses proved prognostic accuracy. Then, its prognostic value was validated in seven external validation series. A nomogram was constructed to guide the individualized treatment of patients with lung adenocarcinoma. Biological pathway and tumor immune infiltration analysis showed that the signature might cause poor prognosis by blocking NK cell activation. Finally, the signature in our centers was confirmed by real-time quantitative PCR. A robust ISGs-related feature was discovered to effectively classify pan-cancer patients into subgroups with different OS.

PTML Model for Selection of Nanoparticles, Anticancer Drugs, and Vitamins in the Design of Drug-Vitamin Nanoparticle Release Systems for Cancer Cotherapy.

Nanosystems are gaining momentum in pharmaceutical sciences because of the wide variety of possibilities for designing these systems to have specific functions. Specifically, studies of new cancer cotherapy drug-vitamin release nanosystems (DVRNs) including anticancer compounds and vitamins or vitamin derivatives have revealed encouraging results. However, the number of possible combinations of design and synthesis conditions is remarkably high. In addition, a large number of anticancer and vitamin derivatives have been already assayed, but a notably less number of cases of DVRNs were assayed as a whole (with the anticancer compound and the vitamin linked to them). Our approach combines with the perturbation theory and machine learning (PTML) model to predict the probability of obtaining an interesting DVRN by changing the anticancer compound and/or the vitamin present in a DVRN that is already tested for other anticancer compounds or vitamins that have not been tested yet as part of a DVRN. In a previous work, we built a linear PTML model useful for the design of these nanosystems. In doing so, we used information fusion (IF) techniques to carry out data enrichment of DVRN data compiled from the literature with the data for preclinical assays of vitamins from the ChEMBL database. The design features of DVRNs and the assay conditions of nanoparticles (NPs) and vitamins were included as multiplicative PT operators (PTOs) to the system, which indicates the importance of these variables. However, the previous work omitted experiments with nonlinear ML techniques and different types of PTOs such as metric-based PTOs. More importantly, the previous work does not consider the structure of the anticancer drug to be included in the new DVRNs. In this work, we are going to accomplish three main objectives (tasks). In the first task, we found a new model, alternative to the one published before, for the rational design of DVRNs using metric-based PTOs. The most accurate PTML model was the artificial neural network model, which showed values of specificity, sensitivity, and accuracy in the range of 90-95% in training and external validation series for more than 130,000 cases (DVRNs vs ChEMBL assays). Furthermore, in the second task, we used IF techniques to carry out data enrichment of our previous data set. In doing so, we constructed a new working data set of >970,000 cases with the data of preclinical assays of DVRNs, vitamins, and anticancer compounds from the ChEMBL database. All these assays have multiple continuous variables or descriptors dk and categorical variables cj (conditions of the assays) for drugs (dack, cacj), vitamins (dvk, cvj), and NPs (dnk, cnj). These data include >20,000 potential anticancer compounds with >270 protein targets (cac1), >580 assay cell organisms (cac2), and so forth. Furthermore, we include >36,000 assay vitamin derivatives in >6200 types of cells (c2vit), >120 assay organisms (c3vit), >60 assay strains (c4vit), and so forth. The enriched data set also contains >20 types of DVRNs (c5n) with 9 NP core materials (c4n), 8 synthesis methods (c7n), and so forth. We expressed all this information with PTOs and developed a qualitatively new PTML model that incorporates information of the anticancer drugs. This new model presents 96-97% of accuracy for training and external validation subsets. In the last task, we carried out a comparative study of ML and/or PTML models published and described how the models we are presenting cover the gap of knowledge in terms of drug delivery. In conclusion, we present here for the first time a multipurpose PTML model that is able to select NPs, anticancer compounds, and vitamins and their conditions of assay for DVRN design.

Development and validation of a novel epigenetic signature for predicting prognosis in colon cancer.

Epigenetic factors play a critical role in carcinogenesis by imparting a distinct feature to the chromatin architecture. The present study aimed to develop a novel epigenetic signature for evaluating the relapse-free survival of colon cancer patients. Public microarray datasets were acquired from the Gene Expression Omnibus databases: GSE39582, GSE17538, GSE33113, and GSE37892 set. Patients from GSE39582 set were randomized 1:1 into training and internal validation series. Patients were divided into high-risk and low-risk groups in training series based on a set of 11 epigenetic factors (p < .001). The good reproducibility for the prognostic value of the epigenetic signature was confirmed in the internal validation series (p < .001), external validation series (a combination of GSE17538 set, GSE33113 set, and GSE37892 set; p = .018), and entire series (p < .001). Furthermore, a nomogram, which integrated the epigenetic signature, pathological stage, and postoperative chemotherapy, was developed based on the GSE39582 set. The time-dependent receiver operating characteristic curve at 1 year demonstrated that the comprehensive signature presented superior prognostic value than the pathological stage. In conclusion, an epigenetic signature, which could be utilized to divide colon cancer patients into two groups with significantly different risk of relapse, was established. This biomarker would aid in identifying patients who require an intensive follow-up and aggressive therapeutic intervention.

Big Data Challenges Targeting Proteins in GPCR Signaling Pathways; Combining PTML-ChEMBL Models and [35S]GTPγS Binding Assays.

G-protein-coupled receptors (GPCRs), also known as 7-transmembrane receptors, are the single largest class of drug targets. Consequently, a large amount of preclinical assays having GPCRs as molecular targets has been released to public sources like the Chemical European Molecular Biology Laboratory (ChEMBL) database. These data are also very complex covering changes in drug chemical structure and assay conditions like c0 = activity parameter (Ki, IC50, etc.), c1 = target protein, c2 = cell line, c3 = assay organism, etc., making difficult the analysis of these databases that are placed in the borders of a Big Data challenge. One of the aims of this work is to develop a computational model able to predict new GPCRs targeting drugs taking into consideration multiple conditions of assay. Another objective is to perform new predictive and experimental studies of selective 5-HTA2 receptor agonist, antagonist, or inverse agonist in human comparing the results with those from the literature. In this work, we combined Perturbation Theory (PT) and Machine Learning (ML) to seek a general PTML model for this data set. We analyzed 343 738 unique compounds with 812 072 end points (assay outcomes), with 185 different experimental parameters, 592 protein targets, 51 cell lines, and/or 55 organisms (species). The best PTML linear model found has three input variables only and predicted 56 202/58 653 positive outcomes (sensitivity = 95.8%) and 470 230/550 401 control cases (specificity = 85.4%) in training series. The model also predicted correctly 18 732/19 549 (95.8%) of positive outcomes and 156 739/183 469 (85.4%) of cases in external validation series. To illustrate its practical use, we used the model to predict the outcomes of six different 5-HT2A receptor drugs, namely, TCB-2, DOI, DOB, altanserin, pimavanserin, and nelotanserin, in a very large number of different pharmacological assays. 5-HT2A receptors are altered in schizophrenia and represent drug target for antipsychotic therapeutic activity. The model correctly predicted 93.83% (76 of 86) experimental results for these compounds reported in ChEMBL. Moreover, [35S]GTPγS binding assays were performed experimentally with the same six drugs with the aim of determining their potency and efficacy in the modulation of G-proteins in human brain tissue. The antagonist ketanserin was included as inactive drug with demonstrated affinity for 5-HT2A/C receptors. Our results demonstrate that some of these drugs, previously described as serotonin 5-HT2A receptor agonists, antagonists, or inverse agonists, are not so specific and show different intrinsic activity to that previously reported. Overall, this work opens a new gate for the prediction of GPCRs targeting compounds.

Development and Validation of an Autophagy Score Signature for the Prediction of Post-operative Survival in Colorectal Cancer.

Background: Survival rates for Colorectal cancer (CRC) patients who experienced early relapse have usually been relatively low. Our study aims at developing an autophagy signature that could help to detect early relapse cases in CRC.Methods: Propensity score matching analysis was carried out between patients from the early relapse group and the long-term survival group from GSE39582. For both groups, respectively, global autophagy expression changes were then analyzed to identify the differentially expressed prognostic autophagy related genes by conducting Linear Models for Microarray data method analysis. Then, the multi-gene signature was validated in TCGA and Fudan University Shanghai Cancer Center (FUSCC) cohorts. Time-dependent ROC were used to test the efficiency of this signature feature in predicting the prognosis of CRC patients.Results: 5 autophagy genes were finally identified to build an early relapse classifier. With specific risk score formula, patients were classified into low- or high-risk group. Time-dependent ROC analyses proved its prognostic accuracy, with AUC 0.841 and 0.803 at 1 and 3 years, respectively. Then, we validated its prognostic value in two external validation series (GSE17538 and GSE33113) and proved that the result is indeed significant irrespective of datasets in two external independent validation cohorts (TCGA and FUSCC cohorts). A nomogram was constructed to guide individualized treatment of patients with CRC.Conclusions: The identification of robust autophagy-related features can effectively classify CRC patients into groups with low and high risk of early relapse. This signature may be used to help select high-risk CRC patients who require more aggressive treatment interventions.

Improvement of Epitope Prediction Using Peptide Sequence Descriptors and Machine Learning.

In this work, we improved a previous model used for the prediction of proteomes as new B-cell epitopes in vaccine design. The predicted epitope activity of a queried peptide is based on its sequence, a known reference epitope sequence under specific experimental conditions. The peptide sequences were transformed into molecular descriptors of sequence recurrence networks and were mixed under experimental conditions. The new models were generated using 709,100 instances of pair descriptors for query and reference peptide sequences. Using perturbations of the initial descriptors under sequence or assay conditions, 10 transformed features were used as inputs for seven Machine Learning methods. The best model was obtained with random forest classifiers with an Area Under the Receiver Operating Characteristics (AUROC) of 0.981 ± 0.0005 for the external validation series (five-fold cross-validation). The database included information about 83,683 peptides sequences, 1448 epitope organisms, 323 host organisms, 15 types of in vivo processes, 28 experimental techniques, and 505 adjuvant additives. The current model could improve the in silico predictions of epitopes for vaccine design. The script and results are available as a free repository.

Prognostic and predictive value of an autophagy-related signature for early relapse in stages I-III colon cancer.

We postulated that expression differences of autophagy-related genes are instrumental in stratifying the risk of early relapse after surgery and evaluating the prognosis of patients with stages I-III colon cancer. Therefore, propensity score matching analysis was performed between patients in early relapse group and long-term survival group from GSE39582 test series and internal validation series. Using Cox regression model, a nine-autophagy-related signature (CAPN2, ATG16L2, TP63, SIRT1, RPS6KB1, PEX3, ATG5, UVRAG, NAF1) was established to classify patients into those at high risk of early relapse (high-risk group), and those at low risk of early relapse (low-risk group). Relapse-free survival (RFS) was significantly different between the two groups in test [hazard ratio (HR): 2.019, 95% confidence interval (CI): 1.362-2.992, P < 0.001], internal validation (HR: 2.464, 95% CI: 1.196-5.079, P < 0.001) and another two external validation series (GSE14333-HR: 2.250, 95% CI: 1.227-4.126, P = 0.007; GSE33113-HR: 5.552, 95% CI: 2.098-14.693, P < 0.001). Then, based on RFS, we developed a nomogram, integrating the nine-autophagy-related classifier and four clinicopathological risk factors to evaluate prognosis of stages I-III colon cancer patients. Time-dependent receiver operating curve at 2 years showed that the integrated signature (area under curve = 0.758) had better prognostic accuracy than American Joint Committee on Cancer TNM stage (area under curve = 0.620). In conclusion, we identified and built a nine-autophagy-related signature, a credible approach to early relapse prediction in stages I-III colon cancer patients, which can assist physicians in devising more efficient therapeutic strategies.

PTML Combinatorial Model of ChEMBL Compounds Assays for Multiple Types of Cancer.

Determining the target proteins of new anticancer compounds is a very important task in Medicinal Chemistry. In this sense, chemists carry out preclinical assays with a high number of combinations of experimental conditions (c j). In fact, ChEMBL database contains outcomes of 65 534 different anticancer activity preclinical assays for 35 565 different chemical compounds (1.84 assays per compound). These assays cover different combinations of c j formed from >70 different biological activity parameters ( c0), >300 different drug targets ( c1), >230 cell lines ( c2), and 5 organisms of assay ( c3) or organisms of the target ( c4). It include a total of 45 833 assays in leukemia, 6227 assays in breast cancer, 2499 assays in ovarian cancer, 3499 in colon cancer, 3159 in lung cancer, 2750 in prostate cancer, 601 in melanoma, etc. This is a very complex data set with multiple Big Data features. This data is hard to be rationalized by researchers to extract useful relationships and predict new compounds. In this context, we propose to combine perturbation theory (PT) ideas and machine learning (ML) modeling to solve this combinatorial-like problem. In this work, we report a PTML (PT + ML) model for ChEMBL data set of preclinical assays of anticancer compounds. This is a simple linear model with only three variables. The model presented values of area under receiver operating curve = AUROC = 0.872, specificity = Sp(%) = 90.2, sensitivity = Sn(%) = 70.6, and overall accuracy = Ac(%) = 87.7 in training series. The model also have Sp(%) = 90.1, Sn(%) = 71.4, and Ac(%) = 87.8 in external validation series. The model use PT operators based on multicondition moving averages to capture all the complexity of the data set. We also compared the model with nonlinear artificial neural network (ANN) models obtaining similar results. This confirms the hypothesis of a linear relationship between the PT operators and the classification as anticancer compounds in different combinations of assay conditions. Last, we compared the model with other PTML models reported in the literature concluding that this is the only one PTML model able to predict activity against multiple types of cancer. This model is a simple but versatile tool for the prediction of the targets of anticancer compounds taking into consideration multiple combinations of experimental conditions in preclinical assays.

Perturbation-Theory and Machine Learning (PTML) Model for High-Throughput Screening of Parham Reactions: Experimental and Theoretical Studies.

Machine learning (ML) algorithms are gaining importance in the processing of chemical information and modeling of chemical reactivity problems. In this work, we have developed a perturbation-theory and machine learning (PTML) model combining perturbation theory (PT) and ML algorithms for predicting the yield of a given reaction. For this purpose, we have selected Parham cyclization, which is a general and powerful tool for the synthesis of heterocyclic and carbocyclic compounds. This reaction has both structural (substitution pattern on the substrate, internal electrophile, ring size, etc.) and operational variables (organolithium reagent, solvent, temperature, time, etc.), so predicting the effect of changes on substrate design (internal elelctrophile, halide, etc.) or reaction conditions on the yield is an important task that could help to optimize the reaction design. The PTML model developed uses PT operators to account for perturbations under experimental conditions and/or structural variables of all the molecules involved in a query reaction, compared to a reaction of reference. Thus, a dataset of >100 reactions has been collected for different substrates and internal electrophiles, under different reaction conditions, with a wide range of yields (0-98%). The best PTML model found using General Linear Regression (GLR) has R = 0.88 in training and R = 0.83 in external validation series for 10 000 pairs of query and reference reactions. The PTML model has a final R = 0.95 for all reactions using multiple reactions of reference. We also report a comparative study of linear versus nonlinear PTML models based on artificial neural network (ANN) algorithms. PTML-ANN models (LNN, MLP, RBF) with R ≈ 0.1-0.8 do not outperform the first PMTL model. This result confirms the validity of the linearity of the model. Next, we carried out an experimental and theoretical study of nonreported Parham reactions to illustrate the practical use of the PTML model. A 500 000-point simulation and a Hammett analysis of the reactivity space of Parham reactions are also reported.

Perturbation Theory/Machine Learning Model of ChEMBL Data for Dopamine Targets: Docking, Synthesis, and Assay of New l-Prolyl-l-leucyl-glycinamide Peptidomimetics.

Predicting drug-protein interactions (DPIs) for target proteins involved in dopamine pathways is a very important goal in medicinal chemistry. We can tackle this problem using Molecular Docking or Machine Learning (ML) models for one specific protein. Unfortunately, these models fail to account for large and complex big data sets of preclinical assays reported in public databases. This includes multiple conditions of assays, such as different experimental parameters, biological assays, target proteins, cell lines, organism of the target, or organism of assay. On the other hand, perturbation theory (PT) models allow us to predict the properties of a query compound or molecular system in experimental assays with multiple boundary conditions based on a previously known case of reference. In this work, we report the first PTML (PT + ML) study of a large ChEMBL data set of preclinical assays of compounds targeting dopamine pathway proteins. The best PTML model found predicts 50000 cases with accuracy of 70-91% in training and external validation series. We also compared the linear PTML model with alternative PTML models trained with multiple nonlinear methods (artificial neural network (ANN), Random Forest, Deep Learning, etc.). Some of the nonlinear methods outperform the linear model but at the cost of a notable increment of the complexity of the model. We illustrated the practical use of the new model with a proof-of-concept theoretical-experimental study. We reported for the first time the organic synthesis, chemical characterization, and pharmacological assay of a new series of l-prolyl-l-leucyl-glycinamide (PLG) peptidomimetic compounds. In addition, we performed a molecular docking study for some of these compounds with the software Vina AutoDock. The work ends with a PTML model predictive study of the outcomes of the new compounds in a large number of assays. Therefore, this study offers a new computational methodology for predicting the outcome for any compound in new assays. This PTML method focuses on the prediction with a simple linear model of multiple pharmacological parameters (IC50, EC50, Ki, etc.) for compounds in assays involving different cell lines used, organisms of the protein target, or organism of assay for proteins in the dopamine pathway.

Chromosome Gene Orientation Inversion Networks (GOINs) of Plasmodium Proteome.

The spatial distribution of genes in chromosomes seems not to be random. For instance, only 10% of genes are transcribed from bidirectional promoters in humans, and many more are organized into larger clusters. This raises intriguing questions previously asked by different authors. We would like to add a few more questions in this context, related to gene orientation inversions. Does gene orientation (inversion) follow a random pattern? Is it relevant to biological activity somehow? We define a new kind of network coined as the gene orientation inversion network (GOIN). GOIN's complex network encodes short- and long-range patterns of inversion of the orientation of pairs of gene in the chromosome. We selected Plasmodium falciparum as a case of study due to the high relevance of this parasite to public health (causal agent of malaria). We constructed here for the first time all of the GOINs for the genome of this parasite. These networks have an average of 383 nodes (genes in one chromosome) and 1314 links (pairs of gene with inverse orientation). We calculated node centralities and other parameters of these networks. These numerical parameters were used to study different properties of gene inversion patterns, for example, distribution, local communities, similarity to Erdös-Rényi random networks, randomness, and so on. We find clues that seem to indicate that gene orientation inversion does not follow a random pattern. We noted that some gene communities in the GOINs tend to group genes encoding for RIFIN-related proteins in the proteome of the parasite. RIFIN-like proteins are a second family of clonally variant proteins expressed on the surface of red cells infected with Plasmodium falciparum. Consequently, we used these centralities as input of machine learning (ML) models to predict the RIFIN-like activity of 5365 proteins in the proteome of Plasmodium sp. The best linear ML model found discriminates RIFIN-like from other proteins with sensitivity and specificity 70-80% in training and external validation series. All of these results may point to a possible biological relevance of gene orientation inversion not directly dependent on genetic sequence information. This work opens the gate to the use of GOINs as a tool for the study of the structure of chromosomes and the study of protein function in proteome research.

Seizure‐susceptible brain regions in glioblastoma: identification of patients at risk

The main aim of this study was to identify which patients with glioblastoma multiforme (GBM) have a higher risk of presenting seizures during follow-up. Patients with newly diagnosed GBM were reviewed (n = 306) and classified as patients with (Group 1) and without (Group 2) seizures at onset. Group 2 was split into patients with seizures during follow-up (Group 2A) and patients who never had seizures (Group 2B). The anatomical location of GBM was identified and compared by voxel-based lesion symptom mapping (discovery set). Seizure-susceptible brain regions obtained were assessed visually and automatically in external GBM validation series (n = 85). In patients with GBM who had no seizures at onset, an increased risk of presenting seizures during follow-up was identified in the superior frontal and inferior occipital lobe, as well as in inferoposterior regions of the temporal lobe. Conversely, those patients with GBM located in medial and inferoanterior temporal areas had a significantly lower risk of suffering from seizures during follow-up. Additionally, the seizure-susceptible brain region maps obtained classified patients in the validation set with high positive and negative predictive values. Tumor location is a useful marker to identify patients with GBM who are at risk of suffering from seizures during follow-up. These results may help to support the use of antiepileptic prophylaxis in a selected GBM population and to improve stratification in antiepileptic clinical trials.

PTML Model for Proteome Mining of B-Cell Epitopes and Theoretical-Experimental Study of Bm86 Protein Sequences from Colima, Mexico.

In this work, we developed a general perturbation theory and machine learning method for data mining of proteomes to discover new B-cell epitopes useful for vaccine design. The method predicts the epitope activity εq(cqj) of one query peptide (q-peptide) under a set of experimental query conditions (cqj). The method uses as input the sequence of the q-peptide. The method also uses as input information about the sequence and epitope activity εr(crj) of a peptide of reference (r-peptide) assayed under similar experimental conditions (crj). The model proposed here is able to classify 1 048 190 pairs of query and reference peptide sequences from the proteome of many organisms reported on IEDB database. These pairs have variations (perturbations) under sequence or assay conditions. The model has accuracy, sensitivity, and specificity between 71 and 80% for training and external validation series. The retrieved information contains structural changes in 83 683 peptides sequences (Seq) determined in experimental assays with boundary conditions involving 1448 epitope organisms (Org), 323 host organisms (Host), 15 types of in vivo process (Proc), 28 experimental techniques (Tech), and 505 adjuvant additives (Adj). Afterward, we reported the experimental sampling, isolation, and sequencing of 15 complete sequences of Bm86 gene from state of Colima, Mexico. Last, we used the model to predict the epitope immunogenic scores under different experimental conditions for the 26 112 peptides obtained from these sequences. The model may become a useful tool for epitope selection toward vaccine design. The theoretical-experimental results on Bm86 protein may help the future design of a new vaccine based on this protein.

Multi-Target Mining of Alzheimer Disease Proteome with Hansch's QSBR-Perturbation Theory and Experimental-Theoretic Study of New Thiophene Isosters of Rasagiline.

Hansch's model is a classic approach to Quantitative Structure-Binding Relationships (QSBR) problems in Pharmacology and Medicinal Chemistry. Hansch QSAR equations are used as input parameters of electronic structure and lipophilicity. In this work, we perform a review on Hansch's analysis. We also developed a new type of PT-QSBR Hansch's model based on Perturbation Theory (PT) and QSBR approach for a large number of drugs reported in CheMBL. The targets are proteins expressed by the Hippocampus region of the brain of Alzheimer Disease (AD) patients. The model predicted correctly 49312 out of 53783 negative perturbations (Specificity = 91.7%) and 16197 out of 21245 positive perturbations (Sensitivity = 76.2%) in training series. The model also predicted correctly 49312/53783 (91.7%) and 16197/21245 (76.2%) negative or positive perturbations in external validation series. We applied our model in theoretical-experimental studies of organic synthesis, pharmacological assay, and prediction of unmeasured results for a series of compounds similar to Rasagiline (compound of reference) with potential neuroprotection effect.

A long non-coding RNA signature to improve prognosis prediction of gastric cancer.

BackgroundIncreasing evidence suggests long non-coding RNAs (lncRNAs) are frequently aberrantly expressed in cancers, however, few related lncRNA signatures have been established for prediction of cancer prognosis. We aimed at developing alncRNA signature to improve prognosis prediction of gastric cancer (GC).MethodsUsing a lncRNA-mining approach, we performed lncRNA expression profiling in large GC cohorts from Gene Expression Ominus (GEO), including GSE62254 data set (N = 300) and GSE15459 data set (N = 192). We established a set of 24-lncRNAs that were significantly associated with the disease free survival (DFS) in the test series.ResultsBased on this 24-lncRNA signature, the test series patients could be classified into high-risk or low-risk subgroup with significantly different DFS (HR = 1.19, 95 % CI = 1.13–1.25, P < 0.0001). The prognostic value of this 24-lncRNA signature was confirmed in the internal validation series and another external validation series, respectively. Further analysis revealed that the prognostic value of this signature was independent of lymph node ratio (LNR) and postoperative chemotherapy. Gene set enrichment analysis (GSEA) indicated that high risk score group was associated with several cancer recurrence and metastasis associated pathways.ConclusionsThe identification of the prognostic lncRNAs indicates the potential roles of lncRNAs in GC biogenesis. Our results may provide an efficient classification tool for clinical prognosis evaluation of GC.Electronic supplementary materialThe online version of this article (doi:10.1186/s12943-016-0544-0) contains supplementary material, which is available to authorized users.

Short-term Survival in Acutely Decompensated Cirrhotic Patients

Aim: The present study was aimed at the early identification of the prognostic factors for 30-day mortality in acutely decompensated cirrhotic patients. Methods: Logistic regression models were used to study the predictors of mortality. Variables significant on univariate testing were included for the multivariate analysis. ROC curves were constructed. The model used retrospective data from 228 patients; and was prospectively validated among 64 patients from the Hospital Clinic: internal validation and 90 patients from Hospital Gregorio Maranon: external validation. Results: The model identified age at admission, serum concentrations of bilirubin, creatinine and sodium, and INR obtained 2 to 8 days after admission as predictors of death in this population. The resulting risk score was highly accurate: AUROC: 0.9150, 95%CI: 0.8509-0.9790 also in the internal and external validation series, but not better that the most widely used scores in hepatology: MELD: 0.8335, 95%CI: 0.7486-0.9184, MELD-Na: 0.8565, 95%CI: 0.7774-0.9356, iMELD: 0.8972, 95%CI: 0.8297-0.9648 and MESO Index: 0.8464, 95%CI: 0.7656-0.9272. The cutoff levels: LR+, LR- of the new score, MELD and MELD-Na that best predicted 30 days mortality were -0.09: 38.6, 0.51, 28: 16.7, 0.42 and 47: 12, 0.7, respectively. Conclusions: MELD, as well as new, more complicated and scanty used scores, obtained 2 to 8 days after admission allows the early and easy identification of patients with an acute decompensation of cirrhosis at high-risk of death on short-term follow-up. These scores may represent a useful tool to select the population suitable for studies to evaluate the efficacy of new therapies and stratify patients in randomized trials.

Multiscale Mapping of AIDS in U.S. Countries vs Anti-HIV Drugs Activity with Complex Networks and Information Indices

In this work, we reviewed different aspects about the epidemiology, drugs, targets, chem-bioinformatics, and systems biology methods, related to AIDS/HIV. Next, we developed a new model to predict complex networks of the prevalence of AIDS in U.S. counties taking into consideration the values of Gini coefficients of social income inequality. We also used activity/structure data of anti-HIV drugs in preclinical assays. First, we trained different Artificial Neural Networks (ANNs) using as input Markov and Symmetry information indices of social networks and of molecular graphs. We obtained the data about AIDS prevalence and Gini coefficient from the AIDSVu database of Emory University and the data about anti-HIV drugs from ChEMBL database. We used Box-Jenkins operators to measure the shift with respect to average behavior of counties from states and drugs from reference compounds assayed in a given protocol, target, or organism. To train/validate the model and predict the complex network we needed to analyze 43,249 data points including values of AIDS prevalence in 2310 counties in U.S. vs ChEMBL results for 21,582 unique drugs, 9 viral or human protein targets, 4856 protocols, and 10 possible experimental measures. The best model found was a Linear Neural Network (LNN) with Accuracy, Specificity, Sensitivity, and AUROC above 0.72-0.73 in training and external validation series. The new linear equation was shown to be useful to generate complex network maps of drug activity vs AIDS/HIV epidemiology in U.S. at county level. Keywords: Anti-HIV drugs, AIDS in U.S. at county level, Gini coefficient, Multiscale models, Box-Jenkins moving average operators, Shannon Entropy, indices of neighborhood symmetry.

Mapping chemical structure-activity information of HAART-drug cocktails over complex networks of AIDS epidemiology and socioeconomic data of U.S. counties

Using computational algorithms to design tailored drug cocktails for highly active antiretroviral therapy (HAART) on specific populations is a goal of major importance for both pharmaceutical industry and public health policy institutions. New combinations of compounds need to be predicted in order to design HAART cocktails. On the one hand, there are the biomolecular factors related to the drugs in the cocktail (experimental measure, chemical structure, drug target, assay organisms, etc.); on the other hand, there are the socioeconomic factors of the specific population (income inequalities, employment levels, fiscal pressure, education, migration, population structure, etc.) to study the relationship between the socioeconomic status and the disease. In this context, machine learning algorithms, able to seek models for problems with multi-source data, have to be used. In this work, the first artificial neural network (ANN) model is proposed for the prediction of HAART cocktails, to halt AIDS on epidemic networks of U.S. counties using information indices that codify both biomolecular and several socioeconomic factors. The data was obtained from at least three major sources. The first dataset included assays of anti-HIV chemical compounds released to ChEMBL. The second dataset is the AIDSVu database of Emory University. AIDSVu compiled AIDS prevalence for >2300 U.S. counties. The third data set included socioeconomic data from the U.S. Census Bureau. Three scales or levels were employed to group the counties according to the location or population structure codes: state, rural urban continuum code (RUCC) and urban influence code (UIC). An analysis of >130,000 pairs (network links) was performed, corresponding to AIDS prevalence in 2310 counties in U.S. vs. drug cocktails made up of combinations of ChEMBL results for 21,582 unique drugs, 9 viral or human protein targets, 4856 protocols, and 10 possible experimental measures. The best model found with the original data was a linear neural network (LNN) with AUROC>0.80 and accuracy, specificity, and sensitivity≈77% in training and external validation series. The change of the spatial and population structure scale (State, UIC, or RUCC codes) does not affect the quality of the model. Unbalance was detected in all the models found comparing positive/negative cases and linear/non-linear model accuracy ratios. Using synthetic minority over-sampling technique (SMOTE), data pre-processing and machine-learning algorithms implemented into the WEKA software, more balanced models were found. In particular, a multilayer perceptron (MLP) with AUROC=97.4% and precision, recall, and F-measure >90% was found.