A dose-response model for statistical analysis of chemical genetic interactions in CRISPRi screens.

An important application of CRISPR interference (CRISPRi) technology is for identifying chemical-genetic interactions (CGIs). Discovery of genes that interact with exposure to antibiotics can yield insights to drug targets and mechanisms of action or resistance. The objective is to identify CRISPRi mutants whose relative abundance is suppressed (or enriched) in the presence of a drug when the target protein is depleted, reflecting synergistic behavior. Different sgRNAs for a given target can induce a wide range of protein depletion and differential effects on growth rate. The effect of sgRNA strength can be partially predicted based on sequence features. However, the actual growth phenotype depends on the sensitivity of cells to depletion of the target protein. For essential genes, sgRNA efficiency can be empirically measured by quantifying effects on growth rate. We observe that the most efficient sgRNAs are not always optimal for detecting synergies with drugs. sgRNA efficiency interacts in a non-linear way with drug sensitivity, producing an effect where the concentration-dependence is maximized for sgRNAs of intermediate strength (and less so for sgRNAs that induce too much or too little target depletion). To capture this interaction, we propose a novel statistical method called CRISPRi-DR (for Dose-Response model) that incorporates both sgRNA efficiencies and drug concentrations in a modified dose-response equation. We use CRISPRi-DR to re-analyze data from a recent CGI experiment in Mycobacterium tuberculosis to identify genes that interact with antibiotics. This approach can be generalized to non-CGI datasets, which we show via an CRISPRi dataset for E. coli growth on different carbon sources. The performance is competitive with the best of several related analytical methods. However, for noisier datasets, some of these methods generate far more significant interactions, likely including many false positives, whereas CRISPRi-DR maintains higher precision, which we observed in both empirical and simulated data.

A dose-response model for statistical analysis of chemical genetic interactions in CRISPRi screens.

An important application of CRISPR interference (CRISPRi) technology is for identifying chemical-genetic interactions (CGIs). Discovery of genes that interact with exposure to antibiotics can yield insights to drug targets and mechanisms of action or resistance. The objective is to identify CRISPRi mutants whose relative abundance is suppressed (or enriched) in the presence of a drug when the target protein is depleted, reflecting synergistic behavior. Different sgRNAs for a given target can induce a wide range of protein depletion and differential effects on growth rate. The effect of sgRNA strength can be partially predicted based on sequence features. However, the actual growth phenotype depends on the sensitivity of cells to depletion of the target protein. For essential genes, sgRNA efficiency can be empirically measured by quantifying effects on growth rate. We observe that the most efficient sgRNAs are not always optimal for detecting synergies with drugs. sgRNA efficiency interacts in a non-linear way with drug sensitivity, producing an effect where the concentration-dependence is maximized for sgRNAs of intermediate strength (and less so for sgRNAs that induce too much or too little target depletion). To capture this interaction, we propose a novel statistical method called CRISPRi-DR (for Dose-Response model) that incorporates both sgRNA efficiencies and drug concentrations in a modified dose-response equation. We use CRISPRi-DR to re-analyze data from a recent CGI experiment in Mycobacterium tuberculosis to identify genes that interact with antibiotics. This approach can be generalized to non-CGI datasets, which we show via an CRISPRi dataset for E. coli growth on different carbon sources. The performance is competitive with the best of several related analytical methods. However, for noisier datasets, some of these methods generate far more significant interactions, likely including many false positives, whereas CRISPRi-DR maintains higher precision, which we observed in both empirical and simulated data.

Dual ICA to extract interacting sets of genes and conditions from transcriptomic data.

One of the challenges in RNA-Seq studies is finding subsets of genes that share a common mechanism of action or are associated with a regulon/pathway. Existing approaches often extract modules that reflect quantitative similarities (such as genes with correlated log-fold-changes) but do not adequately capture biological significance. In this work, we propose the Dual ICA methodology, which provides an agnostic way to extract "interacting modules" composed of sets of genes and conditions that exhibit strong associations. Dual ICA involves performing Independent Component Analysis (ICA) twice, once on the genes and once on the conditions. Using the resulting signal matrices, we extract respective sets of genes and conditions. The interaction between these sets is quantified using the coefficients from a linear regression and significance is determined through the Wald test and Z-score filtering. These coefficients are equivalent to the outer product of independent components obtained from the two signal matrices. Not only do the gene sets extracted align with known regulons, but the significant interacting modules they instantiate also encompass conditions that influence the expression of these regulons through shared mechanisms of action. Compared to traditional unsupervised clustering methods, Dual ICA demonstrates superior performance and provides explicit gene-condition sets for exploring functional relationships.

Modeling Site-Specific Nucleotide Biases Affecting Himar1 Transposon Insertion Frequencies in TnSeq Data Sets.

TnSeq is a widely used methodology for determining gene essentiality, conditional fitness, and genetic interactions in bacteria. The Himar1 transposon is restricted to insertions at TA dinucleotides, but otherwise, few site-specific biases have been identified. As a result, most analytical approaches assume that insertions are expected to be randomly distributed among TA sites in nonessential regions. However, through analysis of Himar1 transposon libraries in Mycobacterium tuberculosis, we demonstrate that there are site-specific biases that affect the frequency of insertion of the Himar1 transposon at different TA sites. We use machine learning and statistical models to characterize patterns in the nucleotides surrounding TA sites that correlate with high or low insertion counts. We then develop a quantitative model based on these patterns that can be used to predict the expected counts at each TA site based on nucleotide context, which can explain up to half of the variance in insertion counts. We show that these insertion preferences exist in Himar1 TnSeq data sets from other mycobacterial and nonmycobacterial species. We present an improved method for identification of essential genes, called TTN-Fitness, that can better distinguish true biological fitness effects by comparing observed counts to expected counts based on our site-specific model of insertion preferences. Compared to previous essentiality methods, TTN-Fitness can make finer distinctions among genes whose disruption causes a fitness defect (or advantage), separating them out from the large pool of nonessentials, and is able to classify many smaller genes (with few TA sites) that were previously characterized as uncertain. IMPORTANCE When using the Himar1 transposon to create transposon insertion mutant libraries, it is known that the transposon is restricted to insertions at TA dinucleotide sites throughout the genome, and the absence of insertions is used to infer which genes are essential (or conditionally essential) in a bacterial organism. It is widely assumed that insertions in nonessential regions are otherwise random, and this assumption is used as the basis of several methods for statistical analysis of TnSeq data. In this paper, we show that the nucleotide sequence surrounding TA sites influences the magnitude of insertions, and these Himar1 insertion preferences (sequence biases) can partially explain why some sites have higher counts than others. We use this predictive model to make improved estimates of the fitness effects of genes, which help make finer distinctions of the phenotype and biological consequences of disruption of nonessential genes.

Preserved network functional connectivity underlies cognitive reserve in multiple sclerosis.

Cognitive reserve is one's mental resilience or resistance to the effects of structural brain damage. Reserve effects are well established in people with multiple sclerosis (PwMS) and Alzheimer's disease, but the neural basis of this phenomenon is unclear. We aimed to investigate whether preservation of functional connectivity explains cognitive reserve. Seventy-four PwMS and 29 HCs underwent neuropsychological assessment and 3 T MRI. Structural damage measures included gray matter (GM) atrophy and network white matter (WM) tract disruption between pairs of GM regions. Resting-state functional connectivity was also assessed. PwMS exhibited significantly impaired cognitive processing speed (t = 2.14, p = .037) and visual/spatial memory (t = 2.72, p = .008), and had significantly greater variance in functional connectivity relative to HCs within relevant networks (p < .001, p < .001, p = .016). Higher premorbid verbal intelligence, a proxy for cognitive reserve, predicted relative preservation of functional connectivity despite accumulation of GM atrophy (standardized-ß = .301, p = .021). Furthermore, preservation of functional connectivity attenuated the impact of structural network WM tract disruption on cognition (ß = -.513, p = .001, for cognitive processing speed; ß = -.209, p = .066, for visual/spatial memory). The data suggests that preserved functional connectivity explains cognitive reserve in PwMS, helping to maintain cognitive capacity despite structural damage.

Lower self-report fatigue in multiple sclerosis is associated with localized white matter tract disruption between amygdala, temporal pole, insula, and other connected structures.

BACKGROUND: Fatigue, a frequent and disabling symptom for people with multiple sclerosis (PwMS), inconsistently correlates with white matter (WM) pathology. Network-based analysis, accounting for the manner in which lesions disrupt networks of structurally connected gray matter (GM) regions, may provide additional insight. OBJECTIVE: To identify patterns of WM tract disruption which explain self-reported fatigue severity in PwMS. METHODS: 137 PwMS and 50 age- and sex-matched healthy controls (HC) underwent fatigue assessment and brain MRI. Lesion maps were applied to determine the severity of WM tract disruption between pairs of GM regions. Then, the Network-Based-Statistics tool was applied to identify structural networks whose disruption explained fatigue severity. To determine whether these networks explain unique variance above conventional MRI measures and depression, regressions were applied controlling for age, sex, brain volume, T2-lesion volume, and depression. RESULTS: Patient-perceived fatigue in PwMS was positively associated with overall lesion burden (ß = 0.563, p-value < 0.001). In contrast, localized disruptions in WM tracts between regions including the amygdala, insula, hippocampus, putamen, temporal pole, caudal-middle-frontal gyrus, rostral-middle-frontal gyrus, inferior-parietal gyrus, and banks of the superior temporal sulcus were significantly negatively correlated with fatigue in PwMS (ß = -0.586, p-value < 0.001). Average disruption within this specific, localized network explained significant additional variance in fatigue above what was otherwise explained by depression and conventional MRI measures of neuropathology (ΔR2 = 0.078, p-value < 0.001). CONCLUSION: Although overall lesion burden correlates positively with fatigue in PwMS, localized WM damage between the amygdala, temporal pole, and other connected structures is associated with lower severity of patient-perceived fatigue.

White matter tract network disruption explains reduced conscientiousness in multiple sclerosis.

Quantifying white matter (WM) tract disruption in people with multiple sclerosis (PwMS) provides a novel means for investigating the relationship between defective network connectivity and clinical markers. PwMS exhibit perturbations in personality, where decreased Conscientiousness is particularly prominent. This trait deficit influences disease trajectory and functional outcomes such as work capacity. We aimed to identify patterns of WM tract disruption related to decreased Conscientiousness in PwMS. Personality assessment and brain MRI were obtained in 133 PwMS and 49 age- and sex-matched healthy controls (HC). Lesion maps were applied to determine the severity of WM tract disruption between pairs of gray matter regions. Next, the Network-Based-Statistics tool was applied to identify structural networks whose disruption negatively correlates with Conscientiousness. Finally, to determine whether these networks explain unique variance above conventional MRI measures and cognition, regression models were applied controlling for age, sex, brain volume, T2-lesion volume, and cognition. Relative to HCs, PwMS exhibited lower Conscientiousness and slowed cognitive processing speed (p = .025, p = .006). Lower Conscientiousness in PwMS was significantly associated with WM tract disruption between frontal, frontal-parietal, and frontal-cingulate pathways in the left (p = .02) and right (p = .01) hemisphere. The mean disruption of these pathways explained unique additive variance in Conscientiousness, after accounting for conventional MRI markers of pathology and cognition (ΔR2  = .049, p = .029). Damage to WM tracts between frontal, frontal-parietal, and frontal-cingulate cortical regions is significantly correlated with reduced Conscientiousness in PwMS. Tract disruption within these networks explains decreased Conscientiousness observed in PwMS as compared with HCs.