Carl M. Kadie

Dependency networks for inference, collaborative filtering, and data visualization

We describe a graphical model for probabilistic relationships--an alternative to the Bayesian network--called a dependency network. The graph of a dependency network, unlike a Bayesian network, is potentially cyclic. The probability component of a dependency network, like a Bayesian network, is a set of conditional distributions, one for each node given its parents. We identify several basic properties of this representation and describe a computationally efficient procedure for learning the graph and probability components from data. We describe the application of this representation to probabilistic inference, collaborative filtering (the task of predicting preferences), and the visualization of acausal predictive relationships.

FaST linear mixed models for genome-wide association studies

We describe factored spectrally transformed linear mixed models (FaST-LMM), an algorithm for genome-wide association studies (GWAS) that scales linearly with cohort size in both run time and memory use. On Wellcome Trust data for 15,000 individuals, FaST-LMM ran an order of magnitude faster than current efficient algorithms. Our algorithm can analyze data for 120,000 individuals in just a few hours, whereas current algorithms fail on data for even 20,000 individuals (http://mscompbio.codeplex.com/).

Models of attention in computing and communication: from principles to applications

Creating computing and communication systems that sense and reason about human attention by fusing together information from multiple streams.

HIV-1 adaptation to NK-cell-mediated immune pressure

Natural killer (NK) cells have an important role in the control of viral infections, recognizing virally infected cells through a variety of activating and inhibitory receptors. Epidemiological and functional studies have recently suggested that NK cells can also contribute to the control of HIV-1 infection through recognition of virally infected cells by both activating and inhibitory killer immunoglobulin-like receptors (KIRs). However, it remains unknown whether NK cells can directly mediate antiviral immune pressure in vivo in humans. Here we describe KIR-associated amino-acid polymorphisms in the HIV-1 sequence of chronically infected individuals, on a population level. We show that these KIR-associated HIV-1 sequence polymorphisms can enhance the binding of inhibitory KIRs to HIV-1-infected CD4+ T cells, and reduce the antiviral activity of KIR-positive NK cells. These data demonstrate that KIR-positive NK cells can place immunological pressure on HIV-1, and that the virus can evade such NK-cell-mediated immune pressure by selecting for sequence polymorphisms, as was previously described for virus-specific T cells and neutralizing antibodies. NK cells might therefore have a previously underappreciated role in contributing to viral evolution.

Founder Effects in the Assessment of HIV Polymorphisms and HLA Allele Associations

Escape from T cell–mediated immune responses affects the ongoing evolution of rapidly evolving viruses such as HIV. By applying statistical approaches that account for phylogenetic relationships among viral sequences, we show that viral lineage effects rather than immune escape often explain apparent human leukocyte antigen (HLA)–mediated immune-escape mutations defined by older analysis methods. Phylogenetically informed methods identified immune-susceptible locations with greatly improved accuracy, and the associations we identified with these methods were experimentally validated. This approach has practical implications for understanding the impact of host immunity on pathogen evolution and for defining relevant variants for inclusion in vaccine antigens.

Marked Epitope- and Allele-Specific Differences in Rates of Mutation in Human Immunodeficiency Type 1 (HIV-1) Gag, Pol, and Nef Cytotoxic T-Lymphocyte Epitopes in Acute/Early HIV-1 Infection

ABSTRACT During acute human immunodeficiency virus type 1 (HIV-1) infection, early host cellular immune responses drive viral evolution. The rates and extent of these mutations, however, remain incompletely characterized. In a cohort of 98 individuals newly infected with HIV-1 subtype B, we longitudinally characterized the rates and extent of HLA-mediated escape and reversion in Gag, Pol, and Nef using a rational definition of HLA-attributable mutation based on the analysis of a large independent subtype B data set. We demonstrate rapid and dramatic HIV evolution in response to immune pressures that in general reflect established cytotoxic T-lymphocyte (CTL) response hierarchies in early infection. On a population level, HLA-driven evolution was observed in ∼80% of published CTL epitopes. Five of the 10 most rapidly evolving epitopes were restricted by protective HLA alleles (HLA-B*13/B*51/B*57/B*5801; P = 0.01), supporting the importance of a strong early CTL response in HIV control. Consistent with known fitness costs of escape, B*57-associated mutations in Gag were among the most rapidly reverting positions upon transmission to non-B*57-expressing individuals, whereas many other HLA-associated polymorphisms displayed slow or negligible reversion. Overall, an estimated minimum of 30% of observed substitutions in Gag/Pol and 60% in Nef were attributable to HLA-associated escape and reversion events. Results underscore the dominant role of immune pressures in driving early within-host HIV evolution. Dramatic differences in escape and reversion rates across codons, genes, and HLA restrictions are observed, highlighting the complexity of viral adaptation to the host immune response.

Improved linear mixed models for genome-wide association studies

to determine these similarities1. Here, however, we show theoretically and experimentally that carefully selecting a small number of SNPs systematically increases power (that is, it jointly reduces false positives and false negatives), improves calibration (lessens inflation or deflation of the test statistic) and reduces computational cost. Our approach is motivated by two considerations. First, an LMM with no fixed effects using genetic similarities constructed from a set of SNPs is mathematically equivalent to a linear regression of the SNPs on the phenotype (with weights integrated over independent normal distributions having the same variance—in particular, the genetic variance)3. That is, an LMM using a given set of SNPs for genetic similarity is equivalent to (Bayesian) linear regression using those SNPs as covariates to correct for confounding. In theory, this equivalence holds only for certain forms of genetic similarity matrices, such as the realized relationship matrix2,3. In practice, however, the realized relationship matrix and other measures of similarity, such as identity by state1, yield very similar measures of association (Supplementary Note 1), and thus our demonstration is quite general. Second, regardless of the form of regression used for GWAS, the significance of SNP-phenotype association should be determined by conditioning on exactly those SNPs that are associated with the phenotype. These SNPs include causal SNPs, or those nearby that tag causal SNPs, and SNPs that are associated by way of confounding (for example, because of population structure). By conditioning on causal or tagging SNPs, we reduce the noise in the assessment of the association4. By conditioning on SNPs associated because of confounding, we control for such confounding5. Moreover, if a SNP is unrelated to the phenotype, it should not be in the conditioning set. In the particular case in which we use Bayesian linear regression for GWAS, the inclusion of unrelated SNPs in the genetic similarity matrix decreases the relative influence of each SNP on the phenotype (because all SNP weights share the same prior distribution whose variance—the genetic variance in the LMM view—is estimated from the data). The decrease in influence leads to incomplete correction for confounding and hence inflated test statistics and reduced power. We refer to this phenomenon as ‘dilution.’ To identify SNPs that satisfy these principles, we developed a simple heuristic that yields improved power and calibration. First, we order SNPs by their linear-regression P values from lowest to highest. Then we construct genetic similarity matrices with an increasing number of SNPs as previously ordered until we find the first minimum in lGC (the genomic control factor). In practice, the number of SNPs selected is typically smaller than the number of individuals analyzed, a condition that can be exploited by an existing algorithm, FaST-LMM, to yield large computational savings2. The equivalence between the LMM and Bayesian linear regression also implies that, when a given SNP is being tested, that SNP should be excluded from the computation of genetic similarity to avoid using it as a covariate. Including the SNP would make the log likelihood of the null model higher than it should be and lead to deflation of the test statistic and loss of power. We call this phenomenon ‘proximal contamination’. In addition to the SNP being tested, we also exclude those SNPs in close proximity (for example, within 2 centimorgans), as linkage disequilibrium will lead to a similar deflation and loss of power. A naive algorithm for excluding these from the similarity matrix is computationally expensive, so we developed a speedup (Supplementary Note 2). Together, the linear-regression scan to select SNPs for inclusion in the matrix and Supplementary Table 4). Many proteins were either overrepresented or underrepresented in each of the protease data sets, and clustering showed that enzyme specificity had the most influence on the results. Some examples within the top 1,000 proteins showed that for specific proteins, one protease outperformed all the others (Fig. 1c and Supplementary Fig. 3). Our data demonstrated that quantitation based on both spectral counting and peptide intensity was indeed biased when solely relying on a single protease, and this bias affected even the most abundant proteins, sometimes by more than a factor of 1,000. Amino acid analysis revealed that proteins overrepresented in a data set obtained by a particular protease contained relatively more cleavage-specific residues for that protease (Supplementary Fig. 3). Our data stresses that the best proteotypic peptides are not necessarily tryptic, a finding that may affect other quantitative assays such as selected reaction monitoring as well. Raw and processed mass spectrometry identification data are available through thegpm.org at ftp://ftp.proteomecentral.org/ public/0/ice.0.e.

Evidence of Differential HLA Class I-Mediated Viral Evolution in Functional and Accessory/Regulatory Genes of HIV-1

Despite the formidable mutational capacity and sequence diversity of HIV-1, evidence suggests that viral evolution in response to specific selective pressures follows generally predictable mutational pathways. Population-based analyses of clinically derived HIV sequences may be used to identify immune escape mutations in viral genes; however, prior attempts to identify such mutations have been complicated by the inability to discriminate active immune selection from virus founder effects. Furthermore, the association between mutations arising under in vivo immune selection and disease progression for highly variable pathogens such as HIV-1 remains incompletely understood. We applied a viral lineage-corrected analytical method to investigate HLA class I-associated sequence imprinting in HIV protease, reverse transcriptase (RT), Vpr, and Nef in a large cohort of chronically infected, antiretrovirally naïve individuals. A total of 478 unique HLA-associated polymorphisms were observed and organized into a series of “escape maps,” which identify known and putative cytotoxic T lymphocyte (CTL) epitopes under selection pressure in vivo. Our data indicate that pathways to immune escape are predictable based on host HLA class I profile, and that epitope anchor residues are not the preferred sites of CTL escape. Results reveal differential contributions of immune imprinting to viral gene diversity, with Nef exhibiting far greater evidence for HLA class I-mediated selection compared to other genes. Moreover, these data reveal a significant, dose-dependent inverse correlation between HLA-associated polymorphisms and HIV disease stage as estimated by CD4+ T cell count. Identification of specific sites and patterns of HLA-associated polymorphisms across HIV protease, RT, Vpr, and Nef illuminates regions of the genes encoding these products under active immune selection pressure in vivo. The high density of HLA-associated polymorphisms in Nef compared to other genes investigated indicates differential HLA class I-driven evolution in different viral genes. The relationship between HLA class I-associated polymorphisms and lower CD4+ cell count suggests that immune escape correlates with disease status, supporting an essential role of maintenance of effective CTL responses in immune control of HIV-1. The design of preventative and therapeutic CTL-based vaccine approaches could incorporate information on predictable escape pathways.

Central Role of Reverting Mutations in HLA Associations with Human Immunodeficiency Virus Set Point

ABSTRACT Much uncertainty still exists over what T-cell responses need to be induced by an effective human immunodeficiency virus (HIV) vaccine. Previous studies have hypothesized that the effective CD8+ T-cell responses are those driving the selection of escape mutations that reduce viral fitness and therefore revert posttransmission. In this study, we adopted a novel approach to define better the role of reverting escape mutations in immune control of HIV infection. This analysis of sequences from 710 study subjects with chronic C-clade HIV type 1 infection demonstrates the importance of mutations that impose a fitness cost in the control of viremia. Consistent with previous studies, the viral set points associated with each HLA-B allele are strongly correlated with the number of Gag-specific polymorphisms associated with the relevant HLA-B allele (r = −0.56, P = 0.0034). The viral set points associated with each HLA-C allele were also strongly correlated with the number of Pol-specific polymorphisms associated with the relevant HLA-C allele (r = −0.67, P = 0.0047). However, critically, both these correlations were dependent solely on the polymorphisms identified as reverting. Therefore, despite the inevitable evolution of viral escape, viremia can be controlled through the selection of mutations that are detrimental to viral fitness. The significance of these results is in highlighting the rationale for an HIV vaccine that can induce these broad responses.

HLA Class I-Driven Evolution of Human Immunodeficiency Virus Type 1 Subtype C Proteome: Immune Escape and Viral Load

ABSTRACT Human immunodeficiency virus type 1 (HIV-1) mutations that confer escape from cytotoxic T-lymphocyte (CTL) recognition can sometimes result in lower viral fitness. These mutations can then revert upon transmission to a new host in the absence of CTL-mediated immune selection pressure restricted by the HLA alleles of the prior host. To identify these potentially critical recognition points on the virus, we assessed HLA-driven viral evolution using three phylogenetic correction methods across full HIV-1 subtype C proteomes from a cohort of 261 South Africans and identified amino acids conferring either susceptibility or resistance to CTLs. A total of 558 CTL-susceptible and -resistant HLA-amino acid associations were identified and organized into 310 immunological sets (groups of individual associations related to a single HLA/epitope combination). Mutations away from seven susceptible residues, including four in Gag, were associated with lower plasma viral-RNA loads (q < 0.2 [where q is the expected false-discovery rate]) in individuals with the corresponding HLA alleles. The ratio of susceptible to resistant residues among those without the corresponding HLA alleles varied in the order Vpr > Gag > Rev > Pol > Nef > Vif > Tat > Env > Vpu (Fishers exact test; P ≤ 0.0009 for each comparison), suggesting the same ranking of fitness costs by genes associated with CTL escape. Significantly more HLA-B (χ2; P = 3.59 × 10−5) and HLA-C (χ2; P = 4.71 × 10−6) alleles were associated with amino acid changes than HLA-A, highlighting their importance in driving viral evolution. In conclusion, specific HIV-1 residues (enriched in Vpr, Gag, and Rev) and HLA alleles (particularly B and C) confer susceptibility to the CTL response and are likely to be important in the development of vaccines targeted to decrease the viral load.