Biology subjects

Bauer, D. C.

Publications and source records attributed to Bauer, D. C..

5 recordsLinked to original sources

bioRxiv2026

Evaluating the cross-species transferability and scaling of sequence-to-function predictions in AlphaGenome

Deep learning models that predict molecular phenotypes directly from DNA sequence offer a powerful framework for interpreting genomic variation. Recently, AlphaGenome was introduced as a deep sequence-to-function architecture capable of predicting observations that historically required experiments. While the model has shown high accuracy, it was primarily evaluated on human variants scored against a reference genome. Here, we test performance on mouse data, the other species AlphaGenome was trained on although with fivefold fewer features than human (1,128 versus 5,930). We demonstrate that AlphaGenomes predictive performance varies considerably depending on the functional task. Specifically, predicted quantitative expression effects are directionally weak and compressed roughly 100-fold relative to empirical benchmarks across both reconstructed-haplotype and single-variant regimes. In contrast, canonical splice-site disruptions are recognized with near-identical accuracy in mouse and human (AUC 0.96 versus 0.98), displaying no cross-species divergence in predicted effect magnitude. We developed a scoring-approach for AI-agents to autonomously assess AlphaGenome prediction confidence and accurately differentiate between AlphaGenomes robust sequence-level recognition across species and its current limitations when interpreting un-fine-mapped regulatory variants. This demonstrates how GenAI innovations that are still under development can safely be harnessed by wrapping a responsible AI layer around the call to intercept flawed results, thereby adhering to international standards, such as the Australian Voluntary AI Safety Standard (VAISS).

bioinformatics↗

bioRxiv2019

TRIBES: A user-friendly pipeline for relatedness detection and disease gene discovery

SummaryTRIBES is a user-friendly pipeline for relatedness detection in genomic data. TRIBES is the first tool which is both accurate up to 7th degree relatives (e.g. third cousins) and combines essential data processing steps into a single user-friendly pipeline. Furthermore, using a proof-of-principle cohort comprising amyotrophic lateral sclerosis cases with known relationship structures and a known causal mutation in SOD1, we demonstrated that TRIBES can successfully uncover disease susceptibility loci. TRIBES has multiple applications in addition to disease gene mapping, including sample quality control in genome wide association studies and avoiding consanguineous unions in family planning.\n\nAvailabilityTRIBES is freely available on GitHub: https://github.com/aehrc/TRIBES/\n\nContactnatalie.twine@csiro.au\n\nSupplementary informationXXXX

bioinformatics↗

bioRxiv2019

IBD analysis of Australian amyotrophic lateral sclerosis SOD1-mutation carriers identifies five founder events and links sporadic cases to existing ALS families

Amyotrophic lateral sclerosis (ALS) is a neurodegenerative disorder characterised by the loss of upper and lower motor neurons resulting in paralysis and eventual death. Approximately 10% of ALS cases have a family history of disease, while the remaining cases present as apparently sporadic. Heritability studies suggest a significant genetic component to sporadic ALS, and although most sporadic cases have an unknown genetic etiology, some familial ALS mutations have also been found in sporadic cases. This suggests that some sporadic cases may be unrecognised familial cases with reduced disease penetrance. Identifying a familial basis of disease in apparently sporadic ALS cases has significant genetic counselling implications for immediate relatives. A powerful strategy to uncover a familial link is identity-by-descent (IBD) analysis which detects genomic regions that have been inherited from a common ancestor. We performed IBD analysis on 90 Australian familial ALS cases from 25 families and three sporadic ALS cases, each of whom carried one of three SOD1 mutations (p.I114T, p.V149G and p.E101G). We identified five unique haplotypes that carry these mutations in our cohort, indicative of five founder events. This included two different haplotypes that carry SOD1 p.I114T, where one haplotype was present in one sporadic case and 20 families, while the second haplotype was found in the remaining two sporadic cases and one family, thus linking these familial and sporadic cases. Furthermore, we linked two families that carry SOD1 p.V149G and found that SOD1 p.E101G arose independently in each family that carries this mutation.

genetics↗

bioRxiv2019

Host porphobilinogen deaminase deficiency confers malaria resistance in Plasmodium chabaudi but not in Plasmodium falciparum or Plasmodium berghei during intraerythrocytic growth

An important component in host resistance to malaria infection are inherited mutations that give rise to abnormalities and deficiencies in erythrocyte proteins and enzymes. Understanding how such mutations confer protection against the disease may be useful for developing new treatment strategies. A mouse ENU-induced mutagenesis screen for novel malaria resistance-conferring mutations identified a novel nonsense mutation in the gene encoding porphobilinogen deaminase (PBGD) in mice, denoted here as PbgdMRI58155. Heterozygote PbgdMRI58155 mice exhibited approximately 50% reduction in cellular PBGD activity in both mature erythrocytes and reticulocytes, although enzyme activity was approximately 10 times higher in reticulocytes than erythrocytes. When challenged with blood-stage P. chabaudi, which preferentially infects erythrocytes, heterozygote mice showed a modest but significant resistance to infection, including reduced parasite growth. A series of assays conducted to investigate the mechanism of resistance indicated that mutant erythrocyte invasion by P. chabaudi was normal, but that following intraerythrocytic establishment a significantly greater proportions of parasites died and therefore affected their ability to propagate. The Plasmodium resistance phenotype was not recapitulated in Pbgd-deficient mice infected with P. berghei, which prefers reticulocytes, or when P. falciparum was cultured in erythrocytes from patients with acute intermittent porphyria (AIP), which had modest (20-50%) reduced levels of PBGD. Furthermore, the growth of Pbgd-null P. falciparum and Pbgd-null P. berghei parasites, which grew at the same rate as their wild-type counterparts in normal cells, were not affected by the PBGD-deficient background of the AIP erythrocytes or Pbgd-deficient mice. Our results confirm the dispensability of parasite PBGD for P. berghei infection and intraerythrocytic growth of P. falciparum, but for the first time identify a requirement for host erythrocyte PBGD by P. chabaudi during in vivo blood stage infection. IMPORTANCEThe causative agent of malaria, Plasmodium, adopts a parasitic lifestyle during erythrocyte infection, and as such relies on host cell factors for its survival and growth. Host-encoded mutations that alter the availability of these factors confer disease resistance, including several well-known genetic erythrocyte abnormalities that have arisen due to the historical evolutionary pressure of malaria. This study identified in mice a novel malaria resistance-conferring host mutation in the heme biosynthesis enzyme, porphobilinogen deaminase (PBGD), and compared the relative requirements by Plasmodium for the host versus parasite-encoded forms of PBGD in both in vivo and in vitro settings. The findings demonstrated that parasite PBGD was dispensable, but that the host enzyme was important specifically during in vivo infection by P. chabaudi, and collectively suggest that Plasmodium requires a certain threshold of the enzyme to sustain its intraerythrocytic growth. Plasmodium may therefore be vulnerable to other interventions that limit host PBGD activity.

microbiology↗

bioRxiv2018

Unlocking HDR-mediated Nucleotide Editing by identifying high-efficiency target sites using machine learning

Editing individual nucleotides is a crucial component for validating genomic disease association. It currently is hampered by CRISPR-Cas-mediated \"base editing\" being limited to certain nucleotide changes, and only achievable within a small window around CRISPR-Cas target sites. The more versatile alternative, HDR (homology directed repair), has a 4-fold lower efficiency with known optimization factors being largely immutable in experiments. Here, we investigated the variable efficiency-governing factors on a novel mouse dataset using machine learning. We found the sequence composition of the repair template (ssODN) to be a governing factor, where different regions of the ssODN have variable influence, which reflects the underlying biophysical mechanism. Our model improves HDR efficiency by 83% compared to traditionally chosen targets. Using our findings, we develop CUNE (Computational Universal Nucleotide Editor), which enables users to identify and design the optimal targeting strategy using traditional base editing or - for-the-first-time - HDR-mediated nucleotide changes.\n\nCUNE can be run via the web at: https://gt-scan.net/cune

bioinformatics↗