Artificial Intelligence-rationalized balanced PPARα/γ dual agonism resets the dysregulated macrophage processes in inflammatory bowel disease

A computational platform, the Boolean network explorer (BoNE), has recently been developed to infuse AI-enhanced precision into drug discovery; it enables querying and navigating invariant Boolean Implication Networks of disease maps for prioritizing high-value targets. Here we used BoNE to query an Inflammatory Bowel Disease (IBD)-map and prioritize a therapeutic strategy that involves dual agonism of two nuclear receptors, PPARα/γ. Balanced agonism of PPARα/γ was predicted to modulate macrophage processes, ameliorate colitis in network-prioritized animal models, reset the gene expression network from disease to health, and achieve a favorable therapeutic index that tracked other FDA-approved targets. Predictions were validated using a balanced and potent PPARα/γ-dual agonist (PAR5359) in two pre-clinical murine models, i.e., Citrobacter rodentium-induced infectious colitis and DSS-induced colitis. Using a combination of selective inhibitors and agonists, we show that balanced dual agonism promotes bacterial clearance more efficiently than individual agonists, both in vivo and in vitro. PPARα is required and its agonism is sufficient to induce the pro-inflammatory cytokines and cellular ROS, which are essential for bacterial clearance and immunity, whereas PPARγ-agonism blunts these responses, delays microbial clearance and induces the anti-inflammatory cytokine, IL10; balanced dual agonism achieved controlled inflammation while protecting the gut barrier and reversal of the transcriptomic network. Furthermore, dual agonism reversed the defective bacterial clearance observed in PBMCs derived from IBD patients. These findings not only deliver a macrophage modulator for use as barrier-protective therapy in IBD, but also highlight the potential of BoNE to rationalize combination therapy.

Boolean implication analysis unveils candidate universal relationships in microbiome data

Background Microbiomes consist of bacteria, viruses, and other microorganisms, and are responsible for many different functions in both organisms and the environment. Past analyses of microbiomes focused on using correlation to determine linear relationships between microbes and diseases. Weak correlations due to nonlinearity between microbe pairs may cause researchers to overlook critical components of the data. With the abundance of available microbiome, we need a method that comprehensively studies microbiomes and how they are related to each other. Results We collected publicly available datasets from human, environment, and animal samples to determine both symmetric and asymmetric Boolean implication relationships between a pair of microbes. We then found relationships that are potentially invariants, meaning they will hold in any microbe community. In other words, if we determine there is a relationship between two microbes, we expect the relationship to hold in almost all contexts. We discovered that around 330,000 pairs of microbes universally exhibit the same relationship in almost all the datasets we studied, thus making them good candidates for invariants. Our results also confirm known biological properties and seem promising in terms of disease diagnosis. Conclusions Since the relationships are likely universal, we expect them to hold in clinical settings, as well as general populations. If these strong invariants are present in disease settings, it may provide insight into prognostic, predictive, or therapeutic properties of clinically relevant diseases. For example, our results indicate that there is a difference in the microbe distributions between patients who have or do not have IBD, eczema and psoriasis. These new analyses may improve disease diagnosis and drug development in terms of accuracy and efficiency.

Boolean Implication Analysis Unveils Candidate Universal Relationships in Microbiome Data

Background: Microbiomes consist of bacteria, viruses, and other microorganisms, and are responsible for many different functions in both organisms and the environment. Past analyses of microbiomes focused on using correlation to determine linear relationships between microbes and diseases. Weak correlations due to nonlinearity between microbe pairs may cause researchers to overlook critical components of the data. With the abundance of available microbiome, we need a method that comprehensively studies microbiomes and how they are related to each other.Results: We collected publicly available datasets from human, environment, and animal samples to determine both symmetric and asymmetric Boolean implication relationships between a pair of microbes. We then found relationships that are potentially invariants, meaning they will hold in any microbe community. In other words, if we determine there is a relationship between two microbes, we expect the relationship to hold in almost all contexts. We discovered that around 330,000 pairs of microbes universally exhibit the same relationship in almost all the datasets we studied, thus making them good candidates for invariants. Our results also confirm known biological properties and seem promising in terms of disease diagnosis. Conclusions: Since the relationships are likely universal, we expect them to hold in clinical settings, as well as general populations. If these strong invariants are present in disease settings, it may provide insight into prognostic, predictive, or therapeutic properties of clinically relevant diseases. For example, our results indicate that there is a difference in the microbe distributions between patients who have or do not have IBD, eczema and psoriasis. These new analyses may improve disease diagnosis and drug development in terms of accuracy and efficiency.

Boolean Implication Analysis Improves Prediction Accuracy of In Silico Gene Reporting of Retinal Cell Types

The retina is a complex tissue containing multiple cell types that is essential for vision. Understanding the gene expression patterns of various retinal cell types has potential applications in regenerative medicine. Retinal organoids (optic vesicles) derived from pluripotent stem cells have begun to yield insights into the transcriptomics of developing retinal cell types in humans through single cell RNA-sequencing studies. Previous methods of gene reporting have relied upon techniques in vivo using microarray data, or correlational and dimension reduction methods for analyzing single cell RNA-sequencing data in silico. Here, we present a bioinformatic approach using Boolean implication to discover retinal cell type-specific genes. We apply this approach to previously published retina and retinal organoid datasets and improve upon previously published correlational methods. Our method improves the prediction accuracy and reproducibility of marker genes of retinal cell types and discovers several new high confidence cone and rod-specific genes. Furthermore, our method is general and can impact all areas of gene expression analyses in cancer and other human diseases.Significance StatementEfforts to derive retinal cell types from pluripotent stem cells to the end of curing retinal disease require robust characterization of these cell types’ gene expression patterns. The Boolean method described in this study improves prediction accuracy of earlier methods of gene reporting, and allows for the discovery and validation of retinal cell type-specific marker genes. The invariant nature of results from Boolean implication analysis can yield high-value molecular markers that can be used as biomarkers or drug targets.

Boolean Implication Analysis Unveils Candidate Universal Relationships in Microbiome Data

Background: Microbiomes consist of bacteria, viruses, and other microorganisms, and are responsible for many different functions in both organisms and the environment. Some previous analyses of microbiomes focus on the relationships between specific microbiomes and a particular disease. These typically use correlation which is fundamentally symmetric with respect to pairs of microbes. Correlation focuses on the symmetry of the data distribution, and asymmetric data is often discarded as having a weak correlation. With all the data available on the microbiome, there is a need for a method that comprehensively studies microbiomes and how they are related to each other.Results: We collect publicly available datasets from human, environment, and animal samples to determine both symmetric and asymmetric Boolean relationships between a pair of microbes. We then find relationships that are potentially invariants, meaning they will hold in any microbe community. In other words, if we determine there is a relationship between two microbes, we expect the relationship to hold in almost all context. We discovered that certain pairs of microbes always exhibit the same relationship in almost all the datasets we studied, thus making them good candidates for universal relationships. Our results confirm known biological properties and seem promising in terms of disease diagnosis.Conclusions: Since the relationships are likely universal, we expect that they will hold in a clinical setting as well as in the general population. Strong universal relationships may provide insight on prognostic, predictive, or therapeutic properties of a clinically relevant disease. These new analyses may improve disease diagnosis and drug development in terms of accuracy and efficiency.

Boolean Implication Mining for Synthetic Lethal Interactions in AML Identifies Acetyl-CoA Carboxylase As a Synthetic Lethal Partner of the IDH1 mutation

Introduction: Somatic mutations in cancer can directly or indirectly perturb signalling and metabolic pathways that can render a cancer cell susceptible to synthetic lethality. We have developed a novel computational method to accelerate identification of synthetic lethal partners for recurrent mutations in acute myeloid leukemia. Our method is based on the hypothesis that, across multiple cancers, synthetic lethal partners of a mutation will be amplified more frequently or deleted less frequently, with concordant changes in expression, in primary tumor samples harboring the mutation of interest.It uses Boolean implication (if-then rules) mining (Sinha et al, Blood 2015) to efficiently identify candidate synthetic lethal partners of a given mutation. The method is distinct from existing work in that it is not reliant on data collected from cell-lines, which are not biologically equivalent to primary tissue and do not always share the composition of mutations found in vivo, but instead utilizes large pan-cancer primary patient datasets. Pan-cancer analysis discovers robust relationships that are more likely to be independent of cancer subtypes, as well as increases statistical power. Methods: We utilized TCGA data of 12 non-AML cancer data-sets (TCGA Research Network et al, Nat. Gen. 2013) for which recurrent AML mutations were present with a frequency of at least 2.5%. These mutations include Cohesin, IDH1, WT1, KRAS, and RUNX1. Boolean implications (FDR < 0.05) were used to identify genes that have more copies in the presence of a mutation as determined by (i) preferred amplification in the presence of the mutation - if gene B is amplified, then mutation A is present, (ii) deletion only in the absence of the mutation - if mutation A is present, then gene B is not deleted. Next, we remove genes that are passengers in large chromosomal alterations using gene expression filtering. Finally, the resulting gene set is filtered by differential gene expression in AML to yield the set of candidate synthetic lethal (SL) partners for a given mutation in AML. Results: To validate our novel method, we compared our putative SL partners to an independent shRNA library screen (DECIPHER) performed in our laboratory for the IDH1 R132 mutation (mut) expressed in THP-1 cells using a doxycycline-inducible promoter (Chan et al, Nat. Med. 2015). We found 6 out of 29 predicted genes showed synthetic lethality when knocked down in the presence of the mutation (Fisher's exact test, p=0.002) indicating our method could find experimentally confirmed interactions. Interestingly, our method predicted Bcl-w to be a SL partner of IDH1 mut, consistent with the SL interaction we previously described between Bcl-2 family members and IDH1 mut in primary AML. Importantly, we found that acetyl-CoA carboxylase alpha (ACACA), the rate-limiting enzyme that controls lipid biosynthesis, was predicted to be a strong SL partner for IDH1 mut. Selective inhibition of ACACA with independently validated shRNA or the small molecule inhibitors, 5-(tetradecyloxy)-2-furoic acid (TOFA) and Soraphen A, prevented cell proliferation in the presence of IDH1 mut but not with IDH1 wildtype. (R)-2-hydroxyglutarate inhibited oxidative phosphorylation and sensitised cells to ACACA inhibitors suggesting the interaction was mediated through the oncometabolite. Gene expression profiling of IDH1 mut cells indicated upregulation of lipid biogenesis pathways (PHOSPHOLIPID METABOLISM, p=0.001). Furthermore, gene expression of ACACA is higher in primary IDH1 mut samples compared to IDH1 wildtype (p=0.008, fold change = 1.2), and cultured primary IDH1mut blasts show selective sensitisation to ACACA inhibition in vitro (n=5/6 IDH1mut/IDH1 wt, p=0.04). Conclusion: We have developed a computational tool that can predict SL interactions for recurrent mutations in AML, with applicability to other cancers. Our method identified de novo lipogenesis as a critical metabolic pathway linked to a specific mutation and suggests therapeutic inhibition of ACACA with small molecules may be beneficial in IDH1 mut AML. This is consistent with recent understanding of the Warburg effect, which postulates that certain oncogenic mutations may indirectly stimulate macromolecule biosynthesis pathways to promote unrestrained cell growth. Our results indicate that a function of the IDH1 mutation is to inhibit oxidative phosphorylation and stimulate de-novo lipid synthesis. Disclosures Majeti: Forty Seven, Inc.: Consultancy, Equity Ownership, Membership on an entity's Board of Directors or advisory committees.

Formalization of Human Categorization Process Using Interpolative Boolean Algebra

Since the ancient times, it has been assumed that categorization has the basic form of classical sets. This implies that the categorization process rests on the Boolean laws. In the second half of the twentieth century, the classical theory has been challenged in cognitive science. According to the prototype theory, objects belong to categories with intensities, while humans categorize objects by comparing them to prototypes of relevant categories. Such categorization process is governed by the principles of perceived world structure and cognitive economy. Approaching the prototype theory by using truth-functional fuzzy logic has been harshly criticized due to not satisfying the complementation laws. In this paper, the prototype theory is approached by using structure-functional fuzzy logic, the interpolative Boolean algebra. The proposed formalism is within the Boolean frame. Categories are represented as fuzzy sets of objects, while comparisons between objects and prototypes are formalized by using Boolean consistent fuzzy relations. Such relations are directly constructed from a Boolean consistent fuzzy partial order relation, which is treated by Boolean implication. The introduced formalism secures the principles of categorization showing that Boolean laws are fundamental in the categorization process. For illustration purposes, the artificial cognitive system which mimics human categorization activity is proposed.

Profiling of Genetic Switches using Boolean Implications in Expression Data

Summary Correlation analysis assuming coexpression of the genes is a widely used method for gene expression analysis in molecular biology. Yet growing extent, quality and dimensionality of the molecular biological data permits emerging, more sophisticated approaches like Boolean implications. We present an approach which is a combination of the SOM (self organizing maps) machine learning method and Boolean implication analysis to identify relations between genes, metagenes and similarly behaving metagene groups (spots). Our method provides a way to assign Boolean states to genes/metagenes/spots and offers a functional view over significantly variant elements of gene expression data on these three different levels. While being able to cover relations between weakly correlated entities Boolean implication method also decomposes these relations into six implication classes. Our method allows one to validate or identify potential relationships between genes and functional modules of interest and to assess their switching behaviour. Furthermore the output of the method renders it possible to construct and study the network of genes. By providing logical implications as updating rules for the network it can also serve to aid modelling approaches.