Evaluating potential drug targets through human loss-of-function genetic variation

Abstract

Human genetics has informed the clinical development of new drugs, and is beginning to influence the selection of new drug targets. Large-scale DNA sequencing studies have created a catalogue of naturally occurring genetic variants predicted to cause loss of function in human genes, which in principle should provide powerful in vivo models of human genetic "knockouts" to complement model organism knockout studies and inform drug development. Here, we consider the use of predicted loss-of-function (pLoF) variation catalogued in the Genome Aggregation Database (gnomAD) for the evaluation of genes as potential drug targets. Many drug targets, including the targets of highly successful inhibitors such as aspirin and statins, are under natural selection at least as extreme as known haploinsufficient genes, with pLoF variants almost completely depleted from the population. Thus, metrics of gene essentiality should not be used to eliminate genes from consideration as potential targets. The identification of individual humans harboring "knockouts" (biallelic gene inactivation), followed by individual recall and deep phenotyping, is highly valuable to study gene function. In most genes, pLoF alleles are sufficiently rare that ascertainment will be largely limited to heterozygous individuals in outbred populations. Sampling of diverse bottlenecked populations and consanguineous individuals will aid in identification of total "knockouts". Careful filtering and curation of pLoF variants in a gene of interest is necessary in order to identify true LoF individuals for follow-up, and the positional distribution or frequency of true LoF variants may reveal important disease biology. Our analysis suggests that the value of pLoF variant data for drug discovery lies in deep curation informed by the nature of the drug and its indication, as well as the biology of the gene, followed by recall-by-genotype studies in targeted populations. 84.

Transcript expression-aware annotation improves rare variant discovery and interpretation

“…In addition, we provide a precomputed file of the transcript expression value for every possible single nucleotide variant in the human genome. This metric has already proven useful in variant curation for drug target identification (32) and for filtering variants for identification of human knockouts(1). Overall, our metric can be incorporated into variant interpretation in Mendelian disease pipelines, rare variant burden analyses, and the prioritization of variants for recall-by-genotype studies.…”

Prion protein lowering is a disease-modifying therapy across prion disease stages, strains, and endpoints

“…The only established knockout phenotype is a peripheral neuropathy, apparently due to deficiency of myelin maintenance signaling to a Schwann cell receptor 14 , which is histologically evident yet phenotypically mild to undetectable in homozygotes and is not observed in heterozygotes 15,16 . Heterozygous inactivating mutations also appear to be tolerated in humans 17,18 , minimizing any concern about on-target toxicity of pharmacologic PrP lowering.…”

Human loss-of-function variants suggest that partial LRRK2 inhibition is a safe therapeutic strategy for Parkinson’s disease

“…This analysis results in a final dataset of 255 gnomAD individuals and 97 UK biobank individuals with 134 unique high-confidence pLoF variants (Figure 1a), an overall carrier frequency of 0.19%; less than half the frequency estimated from uncurated variants, reaffirming the importance of thorough curation of candidate LoF variants 31 . A subset of 25 gnomAD samples with 19 unique LRRK2 pLoF variants with DNA available were all successfully validated by Sanger sequencing ( Supplementary Table 3).…”

“…However, pLoF variants tend to be rare in human populations 29 , and are also enriched for both sequencing and annotation artefacts 30 . As such, leveraging pLoF variation in drug target assessment typically requires very large collections of genetically and phenotypically characterized individuals, combined with deep curation of the target gene and candidate variants 31 . Although previous studies of pLoF variants in LRRK2 have found no association with PD risk 32 , no study has assessed the broader phenotypic consequences of such variants.…”