arXiv subjects

Daniel R. Schrider

Publications and source records attributed to Daniel R. Schrider.

3 recordsLinked to original sources

arXiv2026

On the limits and opportunities of AI reviewers: Reviewing the reviews of Nature-family papers with 45 expert scientists

With the advancement of AI capabilities, AI reviewers are beginning to be deployed in scientific peer review, yet their capability and credibility remain in question: many scientists simply view them as probabilistic systems without the expertise to evaluate research, while other researchers are more optimistic about their readiness without concrete evidence. Understanding what AI reviewers do well, where they fall short, and what challenges remain is essential. However, existing evaluations of AI reviewers have focused on whether their verdicts match human verdicts (e.g., score alignment, acceptance prediction), which is insufficient to characterize their capabilities and limits. In this paper, we close this gap through a large-scale expert annotation study, in which 45 domain scientists in Physical, Biological, and Health Sciences spent 469 hours rating 2,960 individual criticisms (each targeting one specific aspect of a paper) from human-written and AI-generated reviews of 82 Nature-family papers on correctness, significance, and sufficiency of evidence. On a composite of all three dimensions, a reviewing agent powered by GPT-5.2 scores above each paper's top-rated human reviewer (60.0% vs. 48.2%, p = 0.009), while all three AI reviewers (including Gemini 3.0 Pro and Claude Opus 4.5) exceed the lowest-rated human across every dimension. AI reviewers' accurate criticisms are also more often rated significant and well-evidenced, and surface a distinct 26% of issues no human raises. However, AI reviewers overlap far more than humans do (21% vs. 3% for cross-reviewer pairs), and exhibit 16 recurring weaknesses humans do not share, such as limited subfield knowledge, lack of long context management over multiple files, and overly critical stance on minor issues. Overall, our results position current AI reviewers as complements to, not substitutes for, human reviewers.

cs.CL↗

arXiv2015

Inferring selective constraint from population genomic data suggests recent regulatory turnover in the human brain

The comparative genomics revolution of the past decade has enabled the discovery of functional elements in the human genome via sequence comparison. While that is so, an important class of elements, those specific to humans, is entirely missed by searching for sequence conservation across species. Here we present an analysis based on variation data among human genomes that utilizes a supervised machine learning approach for the identification of human specific purifying selection in the genome. Using only allele frequency information from the complete low coverage 1000 Genomes Project dataset in conjunction with a support vector machine trained from known functional and non-functional portions of the genome, we are able to accurately identify portions of the genome constrained by purifying selection. Our method identifies previously known human-specific gains or losses of function and uncovers many novel candidates. Candidate targets for gain and loss of function along the human lineage include numerous putative regulatory regions of genes essential for normal development of the central nervous system, including a significant enrichment of gain of function events near neurotransmitter receptor genes. These results are consistent with regulatory turnover being a key mechanism in the evolution of human-specific characteristics of brain development. Finally, we show that the majority of the genome is unconstrained by natural selection currently, in agreement with what has been estimated from phylogenetic methods but in sharp contrast to estimates based on transcriptomics or other high-throughput functional methods.

q-bio.PE↗

arXiv2014

Discovering functional DNA elements using population genomic information: A proof of concept using human mtDNA

Identifying the complete set of functional elements within the human genome would be a windfall for multiple areas of biological research including medicine, molecular biology, and evolution. Complete knowledge of function would aid in the prioritization of loci when searching for the genetic bases of disease or adaptive phenotypes. Because mutations that disrupt function are disfavored by natural selection, purifying selection leaves a detectable signature within functional elements; accordingly this signal has been exploited through the use of genomic comparisons of distantly related species. However, the functional complement of the genome changes extensively across time and between lineages, therefore, evidence of the current action of purifying selection is essential. Because the removal of deleterious mutations by natural selection also reduces within-species genetic diversity within functional loci, dense population genetic data have the potential to reveal genomic elements that are currently functional. Here we assess the potential of this approach using 16,411 human mitochondrial genomes. We show that the high density of polymorphism in this dataset precisely delineates regions experiencing purifying selection. Further, we show that the number of segregating alleles at a site is strongly correlated with its divergence across species after accounting for known mutational biases in human mtDNA. These two measures track one another at a remarkably fine scale across many loci--a correlation that is purely the result of natural selection. Our results demonstrate that genetic variation has the potential to reveal exactly which nucleotides in the genome are currently performing important functions and likely to have deleterious fitness effects when mutated. As more complete genomes are sequenced, similar power to reveal purifying selection may be achievable in the human nuclear genome.

q-bio.PE↗