Prediction of disease-free survival for precision medicine using cooperative learning on multi-omic data.

In precision medicine, both predicting the disease susceptibility of an individual and forecasting its disease-free survival are areas of key research. Besides the classical epidemiological predictor variables, data from multiple (omic) platforms are increasingly available. To integrate this wealth of information, we propose new methodology to combine both cooperative learning, a recent approach to leverage the predictive power of several datasets, and polygenic hazard score models. Polygenic hazard score models provide a practitioner with a more differentiated view of the predicted disease-free survival than the one given by merely a point estimate, for instance computed with a polygenic risk score. Our aim is to leverage the advantages of cooperative learning for the computation of polygenic hazard score models via Cox's proportional hazard model, thereby improving the prediction of the disease-free survival. In our experimental study, we apply our methodology to forecast the disease-free survival for Alzheimer's disease (AD) using three layers of data. One layer contains epidemiological variables such as sex, APOE (apolipoprotein E, a genetic risk factor for AD) status and 10 leading principal components. Another layer contains selected genomic loci, and the last layer contains methylation data for selected CpG sites. We demonstrate that the survival curves computed via cooperative learning yield an AUC of around $0.7$, above the state-of-the-art performance of its competitors. Importantly, the proposed methodology returns (1) a linear score that can be easily interpreted (in contrast to machine learning approaches), and (2) a weighting of the predictive power of the involved data layers, allowing for an assessment of the importance of each omic (or other) platform. Similarly to polygenic hazard score models, our methodology also allows one to compute individual survival curves for each patient.

On the effect heterogeneity of established disease susceptibility loci for Alzheimer's disease across different genetic ancestries.

INTRODUCTION: Genome-wide association studies have identified numerous disease susceptibility loci (DSLs) for Alzheimer's disease (AD). However, only a limited number of studies have investigated the dependence of the genetic effect size of established DSLs on genetic ancestry. METHODS: We utilized the whole genome sequencing data from the Alzheimer's Disease Sequencing Project (ADSP) including 35,569 participants. A total of 25,459 subjects in four distinct populations (African ancestry, non-Hispanic White, admixed Hispanic, and Asian) were analyzed. RESULTS: We found that nine DSLs showed significant heterogeneity across populations. Single nucleotide polymorphism (SNP) rs2075650 in translocase of outer mitochondrial membrane 40 (TOMM40) showed the largest heterogeneity (Cochran's Q = 0.00, I2 = 90.08), followed by other SNPs in apolipoprotein C1 (APOC1) and apolipoprotein E (APOE). Two additional loci, signal-induced proliferation-associated 1 like 2 (SIPA1L2) and solute carrier 24 member 4 (SLC24A4), showed significant heterogeneity across populations. DISCUSSION: We observed substantial heterogeneity for the APOE-harboring 19q13.32 region with TOMM40/APOE/APOC1 genes. The largest risk effect was seen among African Americans, while Asians showed a surprisingly small risk effect.

A comparison between similarity matrices for principal component analysis to assess population stratification in sequenced genetic data sets.

Genetic similarity matrices are commonly used to assess population substructure (PS) in genetic studies. Through simulation studies and by the application to whole-genome sequencing (WGS) data, we evaluate the performance of three genetic similarity matrices: the unweighted and weighted Jaccard similarity matrices and the genetic relationship matrix. We describe different scenarios that can create numerical pitfalls and lead to incorrect conclusions in some instances. We consider scenarios in which PS is assessed based on loci that are located across the genome ('globally') and based on loci from a specific genomic region ('locally'). We also compare scenarios in which PS is evaluated based on loci from different minor allele frequency bins: common (>5%), low-frequency (5-0.5%) and rare (<0.5%) single-nucleotide variations (SNVs). Overall, we observe that all approaches provide the best clustering performance when computed based on rare SNVs. The performance of the similarity matrices is very similar for common and low-frequency variants, but for rare variants, the unweighted Jaccard matrix provides preferable clustering features. Based on visual inspection and in terms of standard clustering metrics, its clusters are the densest and the best separated in the principal component analysis of variants with rare SNVs compared with the other methods and different allele frequency cutoffs. In an application, we assessed the role of rare variants on local and global PS, using WGS data from multiethnic Alzheimer's disease data sets and European or East Asian populations from the 1000 Genome Project.

Region-based analysis of rare genomic variants in whole-genome sequencing datasets reveal two novel Alzheimer's disease-associated genes: DTNB and DLG2.

Alzheimer's disease (AD) is a genetically complex disease for which nearly 40 loci have now been identified via genome-wide association studies (GWAS). We attempted to identify groups of rare variants (alternate allele frequency <0.01) associated with AD in a region-based, whole-genome sequencing (WGS) association study (rvGWAS) of two independent AD family datasets (NIMH/NIA; 2247 individuals; 605 families). Employing a sliding window approach across the genome, we identified several regions that achieved association p values <10-6, using the burden test or the SKAT statistic. The genomic region around the dystobrevin beta (DTNB) gene was identified with the burden and SKAT test and replicated in case/control samples from the ADSP study reaching genome-wide significance after meta-analysis (pmeta = 4.74 × 10-8). SKAT analysis also revealed region-based association around the Discs large homolog 2 (DLG2) gene and replicated in case/control samples from the ADSP study (pmeta = 1 × 10-6). In conclusion, in a region-based rvGWAS of AD we identified two novel AD genes, DLG2 and DTNB, based on association with rare variants.

A Smoothed Version of the Lassosum Penalty for Fitting Integrated Risk Models Using Summary Statistics or Individual-Level Data.

Polygenic risk scores are a popular means to predict the disease risk or disease susceptibility of an individual based on its genotype information. When adding other important epidemiological covariates such as age or sex, we speak of an integrated risk model. Methodological advances for fitting more accurate integrated risk models are of immediate importance to improve the precision of risk prediction, thereby potentially identifying patients at high risk early on when they are still able to benefit from preventive steps/interventions targeted at increasing their odds of survival, or at reducing their chance of getting a disease in the first place. This article proposes a smoothed version of the "Lassosum" penalty used to fit polygenic risk scores and integrated risk models using either summary statistics or raw data. The smoothing allows one to obtain explicit gradients everywhere for efficient minimization of the Lassosum objective function while guaranteeing bounds on the accuracy of the fit. An experimental section on both Alzheimer's disease and COPD (chronic obstructive pulmonary disease) demonstrates the increased accuracy of the proposed smoothed Lassosum penalty compared to the original Lassosum algorithm (for the datasets under consideration), allowing it to draw equal with state-of-the-art methodology such as LDpred2 when evaluated via the AUC (area under the ROC curve) metric.

Whole-genome sequencing reveals new Alzheimer's disease-associated rare variants in loci related to synaptic function and neuronal development.

INTRODUCTION: Genome-wide association studies have led to numerous genetic loci associated with Alzheimer's disease (AD). Whole-genome sequencing (WGS) now permits genome-wide analyses to identify rare variants contributing to AD risk. METHODS: We performed single-variant and spatial clustering-based testing on rare variants (minor allele frequency [MAF] ≤1%) in a family-based WGS-based association study of 2247 subjects from 605 multiplex AD families, followed by replication in 1669 unrelated individuals. RESULTS: We identified 13 new AD candidate loci that yielded consistent rare-variant signals in discovery and replication cohorts (4 from single-variant, 9 from spatial-clustering), implicating these genes: FNBP1L, SEL1L, LINC00298, PRKCH, C15ORF41, C2CD3, KIF2A, APC, LHX9, NALCN, CTNNA2, SYTL3, and CLSTN2. DISCUSSION: Downstream analyses of these novel loci highlight synaptic function, in contrast to common AD-associated variants, which implicate innate immunity and amyloid processing. These loci have not been associated previously with AD, emphasizing the ability of WGS to identify AD-associated rare variants, particularly outside of the exome.

Whole-genome sequencing reveals new Alzheimer's disease-associated rare variants in loci related to synaptic function and neuronal development.

INTRODUCTION: Genome-wide association studies have led to numerous genetic loci associated with Alzheimer's disease (AD). Whole-genome sequencing (WGS) now permit genome-wide analyses to identify rare variants contributing to AD risk. METHODS: We performed single-variant and spatial clustering-based testing on rare variants (minor allele frequency ≤1%) in a family-based WGS-based association study of 2,247 subjects from 605 multiplex AD families, followed by replication in 1,669 unrelated individuals. RESULTS: We identified 13 new AD candidate loci that yielded consistent rare-variant signals in discovery and replication cohorts (4 from single-variant, 9 from spatial-clustering), implicating these genes: FNBP1L, SEL1L, LINC00298, PRKCH, C15ORF41, C2CD3, KIF2A, APC, LHX9, NALCN, CTNNA2, SYTL3, CLSTN2. DISCUSSION: Downstream analyses of these novel loci highlight synaptic function, in contrast to common AD-associated variants, which implicate innate immunity. These loci have not been previously associated with AD, emphasizing the ability of WGS to identify AD-associated rare variants, particularly outside of coding regions.

Tau molecular diversity contributes to clinical heterogeneity in Alzheimer's disease.

Alzheimer's disease (AD) causes unrelenting, progressive cognitive impairments, but its course is heterogeneous, with a broad range of rates of cognitive decline1. The spread of tau aggregates (neurofibrillary tangles) across the cerebral cortex parallels symptom severity2,3. We hypothesized that the kinetics of tau spread may vary if the properties of the propagating tau proteins vary across individuals. We carried out biochemical, biophysical, MS and both cell- and animal-based-bioactivity assays to characterize tau in 32 patients with AD. We found striking patient-to-patient heterogeneity in the hyperphosphorylated species of soluble, oligomeric, seed-competent tau. Tau seeding activity correlates with the aggressiveness of the clinical disease, and some post-translational modification (PTM) sites appear to be associated with both enhanced seeding activity and worse clinical outcomes, whereas others are not. These data suggest that different individuals with 'typical' AD may have distinct biochemical features of tau. These data are consistent with the possibility that individuals with AD, much like people with cancer, may have multiple molecular drivers of an otherwise common phenotype, and emphasize the potential for personalized therapeutic approaches for slowing clinical progression of AD.

Identification of Novel Alzheimer's Disease Loci Using Sex-Specific Family-Based Association Analysis of Whole-Genome Sequence Data.

With the advent of whole genome-sequencing (WGS) studies, family-based designs enable sex-specific analysis approaches that can be applied to only affected individuals; tests using family-based designs are attractive because they are completely robust against the effects of population substructure. These advantages make family-based association tests (FBATs) that use siblings as well as parents especially suited for the analysis of late-onset diseases such as Alzheimer's Disease (AD). However, the application of FBATs to assess sex-specific effects can require additional filtering steps, as sensitivity to sequencing errors is amplified in this type of analysis. Here, we illustrate the implementation of robust analysis approaches and additional filtering steps that can minimize the chances of false positive-findings due to sex-specific sequencing errors. We apply this approach to two family-based AD datasets and identify four novel loci (GRID1, RIOK3, MCPH1, ZBTB7C) showing sex-specific association with AD risk. Following stringent quality control filtering, the strongest candidate is ZBTB7C (Pinter = 1.83 × 10-7), in which the minor allele of rs1944572 confers increased risk for AD in females and protection in males. ZBTB7C encodes the Zinc Finger and BTB Domain Containing 7C, a transcriptional repressor of membrane metalloproteases (MMP). Members of this MMP family were implicated in AD neuropathology.

Loss of Ataxin-1 Potentiates Alzheimer's Pathogenesis by Elevating Cerebral BACE1 Transcription.

Expansion of CAG trinucleotide repeats in ATXN1 causes spinocerebellar ataxia type 1 (SCA1), a neurodegenerative disease that impairs coordination and cognition. While ATXN1 is associated with increased Alzheimer's disease (AD) risk, CAG repeat number in AD patients is not changed. Here, we investigated the consequences of ataxin-1 loss of function and discovered that knockout of Atxn1 reduced CIC-ETV4/5-mediated inhibition of Bace1 transcription, leading to increased BACE1 levels and enhanced amyloidogenic cleavage of APP, selectively in AD-vulnerable brain regions. Elevated BACE1 expression exacerbated Aß deposition and gliosis in AD mouse models and impaired hippocampal neurogenesis and olfactory axonal targeting. In SCA1 mice, polyglutamine-expanded mutant ataxin-1 led to the increase of BACE1 post-transcriptionally, both in cerebrum and cerebellum, and caused axonal-targeting deficit and neurodegeneration in the hippocampal CA2 region. These findings suggest that loss of ataxin-1 elevates BACE1 expression and Aß pathology, rendering it a potential contributor to AD risk and pathogenesis.

Silencing of the Drosophila ortholog of SOX5 leads to abnormal neuronal development and behavioral impairment.

SOX5 encodes a transcription factor that is expressed in multiple tissues including heart, lung and brain. Mutations in SOX5 have been previously found in patients with amyotrophic lateral sclerosis (ALS) and developmental delay, intellectual disability and dysmorphic features. To characterize the neuronal role of SOX5, we silenced the Drosophila ortholog of SOX5, Sox102F, by RNAi in various neuronal subtypes in Drosophila. Silencing of Sox102F led to misorientated and disorganized michrochaetes, neurons with shorter dendritic arborization (DA) and reduced complexity, diminished larval peristaltic contractions, loss of neuromuscular junction bouton structures, impaired olfactory perception, and severe neurodegeneration in brain. Silencing of SOX5 in human SH-SY5Y neuroblastoma cells resulted in a significant repression of WNT signaling activity and altered expression of WNT-related genes. Genetic association and meta-analyses of the results in several large family-based and case-control late-onset familial Alzheimer's disease (LOAD) samples of SOX5 variants revealed several variants that show significant association with AD disease status. In addition, analysis for rare and highly penetrate functional variants revealed four novel variants/mutations in SOX5, which taken together with functional prediction analysis, suggests a strong role of SOX5 causing AD in the carrier families. Collectively, these findings indicate that SOX5 is a novel candidate gene for LOAD with an important role in neuronal function. The genetic findings warrant further studies to identify and characterize SOX5 variants that confer risk for AD, ALS and intellectual disability.

Alzheimer's disease-associated TREM2 variants exhibit either decreased or increased ligand-dependent activation.

INTRODUCTION: TREM2 is a lipid-sensing activating receptor on microglia known to be important for Alzheimer's disease (AD), but whether it plays a beneficial or detrimental role in disease pathogenesis is controversial. METHODS: We analyzed AD risk of TREM2 variants in the NIMH AD Genetics Initiative Study and AD Sequencing Project. We compared each variant's risk and functional impact by a reporter assay. Finally, we analyzed expression of TREM2 on human monocytes. RESULTS: We provide more evidence for increased AD risk associated with several TREM2 variants, and show that these variants decreased or markedly increased binding to TREM2 ligands. We identify HDL and LDL as novel TREM2 ligands. We also show that TREM2 expression in human monocytes is minimal compared to monocyte-derived dendritic cells. DISCUSSION: Our results suggest that TREM2 signaling helps protect against AD but can cause harm in excess, supporting the idea that proper TREM2 function is important to counteract disease progression.

Gain-of-function mutations in protein kinase Cα (PKCα) may promote synaptic defects in Alzheimer's disease.

Alzheimer's disease (AD) is a progressive dementia disorder characterized by synaptic degeneration and amyloid-ß (Aß) accumulation in the brain. Through whole-genome sequencing of 1345 individuals from 410 families with late-onset AD (LOAD), we identified three highly penetrant variants in PRKCA, the gene that encodes protein kinase Cα (PKCα), in five of the families. All three variants linked with LOAD displayed increased catalytic activity relative to wild-type PKCα as assessed in live-cell imaging experiments using a genetically encoded PKC activity reporter. Deleting PRKCA in mice or adding PKC antagonists to mouse hippocampal slices infected with a virus expressing the Aß precursor CT100 revealed that PKCα was required for the reduced synaptic activity caused by Aß. In PRKCA(-/-) neurons expressing CT100, introduction of PKCα, but not PKCα lacking a PDZ interaction moiety, rescued synaptic depression, suggesting that a scaffolding interaction bringing PKCα to the synapse is required for its mediation of the effects of Aß. Thus, enhanced PKCα activity may contribute to AD, possibly by mediating the actions of Aß on synapses. In contrast, reduced PKCα activity is implicated in cancer. Hence, these findings reinforce the importance of maintaining a careful balance in the activity of this enzyme.

Adjusting heterogeneous ascertainment bias for genetic association analysis with extended families.

BACKGROUND: In family-based association analysis, each family is typically ascertained from a single proband, which renders the effects of ascertainment bias heterogeneous among family members. This is contrary to case-control studies, and may introduce sample or ascertainment bias. Statistical efficiency is affected by ascertainment bias, and careful adjustment can lead to substantial improvements in statistical power. However, genetic association analysis has often been conducted using family-based designs, without addressing the fact that each proband in a family has had a great influence on the probability for each family member to be affected. METHOD: We propose a powerful and efficient statistic for genetic association analysis that considered the heterogeneity of ascertainment bias among family members, under the assumption that both prevalence and heritability of disease are available. With extensive simulation studies, we showed that the proposed method performed better than the existing methods, particularly for diseases with large heritability. RESULTS: We applied the proposed method to the genome-wide association analysis of Alzheimer's disease. Four significant associations with the proposed method were found. CONCLUSION: Our significant findings illustrated the practical importance of this new analysis method.

The rare TREM2 R47H variant exerts only a modest effect on Alzheimer disease risk.

OBJECTIVES: Recently, 2 independent studies reported that a rare missense variant, rs75932628 (R47H), in exon 2 of the gene encoding the "triggering receptor expressed on myeloid cells 2" (TREM2) significantly increases the risk of Alzheimer disease (AD) with an effect size comparable to that of the APOE ε4 allele. METHODS: In this study, we attempted to replicate the association between rs75932628 and AD risk by directly genotyping rs75932628 in 2 independent Caucasian family cohorts consisting of 927 families (with 1,777 affected and 1,235 unaffected) and in 2 Caucasian case-control cohorts composed of 1,314 cases and 1,609 controls. In addition, we imputed genotypes in 3 independent Caucasian case-control cohorts containing 1,906 cases and 1,503 controls. RESULTS: Meta-analysis of the 2 family-based and the 5 case-control cohorts yielded a p value of 0.0029, while the overall summary estimate (using case-control data only) resulted in an odds ratio of 1.67 (95% confidence interval 0.95-2.92) for the association between the TREM2 R47H and increased AD risk. CONCLUSIONS: While our results serve to confirm the association between R47H and risk of AD, the observed effect on risk was substantially smaller than that previously reported.

Elucidation of the BACE1 regulating factor GGA3 in Alzheimer's disease.

Golgi-localized γ-ear-containing ADP-ribosylation factor-binding protein (GGA3) is a central regulator of trafficking and degradation of BACE1 (ß-site AßPP-cleaving enzyme), the rate-limiting enzyme in the production of amyloid-ß (Aß) in Alzheimer's disease (AD). Here, we assessed the potential role of GGA3 in AD pathogenesis using independent neuropathological, case-control, and family-based human sample cohorts. Increased BACE1 levels coincided with decreased GGA3 levels and with elevated phosphorylation status of eIF2α-Ser51 in the temporal cortex of AD patients as compared to age-matched controls. Severity of the disease did not alter mRNA or protein levels of GGA3 in the inferior temporal cortex of AD patients, while a positive correlation between GGA3 and the levels of total, but not phosphorylated, tau was observed. Genetically, we did not observe consistent evidence for association between AD risk and common GGA3 polymorphisms across a number of independent sample cohorts. However, a nominally significant association was observed with rs2242230 (p < 0.05) among the Finnish case-control cohort. Accordingly, mRNA and protein levels of GGA3 in the inferior temporal cortex of AD patients did not significantly correlate with rs2242230 genotype status. While the present study indicates that GGA3 is involved in the cellular processes relevant for AD pathogenesis, the genetic data do not support the idea that common GGA3 polymorphisms would contribute to AD risk.

Alzheimer's disease risk gene CD33 inhibits microglial uptake of amyloid beta.

The transmembrane protein CD33 is a sialic acid-binding immunoglobulin-like lectin that regulates innate immunity but has no known functions in the brain. We have previously shown that the CD33 gene is a risk factor for Alzheimer's disease (AD). Here, we observed increased expression of CD33 in microglial cells in AD brain. The minor allele of the CD33 SNP rs3865444, which confers protection against AD, was associated with reductions in both CD33 expression and insoluble amyloid beta 42 (Aß42) levels in AD brain. Furthermore, the numbers of CD33-immunoreactive microglia were positively correlated with insoluble Aß42 levels and plaque burden in AD brain. CD33 inhibited uptake and clearance of Aß42 in microglial cell cultures. Finally, brain levels of insoluble Aß42 as well as amyloid plaque burden were markedly reduced in APP(Swe)/PS1(ΔE9)/CD33(-/-) mice. Therefore, CD33 inactivation mitigates Aß pathology and CD33 inhibition could represent a novel therapy for AD.

Potential late-onset Alzheimer's disease-associated mutations in the ADAM10 gene attenuate {alpha}-secretase activity.

ADAM10, a member of a disintegrin and metalloprotease family, is an alpha-secretase capable of anti-amyloidogenic proteolysis of the amyloid precursor protein. Here, we present evidence for genetic association of ADAM10 with Alzheimer's disease (AD) as well as two rare potentially disease-associated non-synonymous mutations, Q170H and R181G, in the ADAM10 prodomain. These mutations were found in 11 of 16 affected individuals (average onset age 69.5 years) from seven late-onset AD families. Each mutation was also found in one unaffected subject implying incomplete penetrance. Functionally, both mutations significantly attenuated alpha-secretase activity of ADAM10 (>70% decrease), and elevated Abeta levels (1.5-3.5-fold) in cell-based studies. In summary, we provide the first evidence of ADAM10 as a candidate AD susceptibility gene, and report two potentially pathogenic mutations with incomplete penetrance for late-onset familial AD.