Search bioRxiv⌕ Search

bioRxiv · 10.64898/2026.11.03.764067

Is the Bitter Lesson coming to genomic prediction? An untuned tabular foundation model matches or beats specialized methods

Abstract

Genomic prediction (GP) is dominated by models designed around assumptions about how genetic variation affects phenotypes. Here we asked how well a general-purpose pretrained tabular foundation model called TabPFN-3.5 does with no genetics-specific modeling and no tuning when given marker genotypes directly. We benchmarked TabPFN-3.5 using a published comparison of 18 GP methods on 18 traits in six plant species. Against the published means, TabPFN-3.5 ranked first on 11 of 18 traits and had the best median rank. Because the published benchmark did not identify its test sets, we reran GBLUP, RKHS and elastic-net models on the same splits as TabPFN-3.5. Against the best of these, TabPFN-3.5 was significantly behind on only one trait, spruce wood density, and ahead on 12 of 18, with significant gains on all three soybean traits and rice flowering time. In the soybean panel, which had the largest number of lines, its advantage was clearest. In leave-families-out cross-validation, where models were trained on some soybean families and used to predict lines from other families withheld from training, every model lost accuracy and TabPFN-3.5 performed on par with GBLUP and RKHS. That a general-purpose learner matches or exceeds decades of specialized GP models is consistent with Sutton's Bitter Lesson, which holds that methods that scale with computation eventually overtake those built on human knowledge of the domain.

Explore related subjects

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Flagel, L.. 2026-11-15. Is the Bitter Lesson coming to genomic prediction? An untuned tabular foundation model matches or beats specialized methods. https://doi.org/10.64898/2026.11.03.764067

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related preprints

Unravelling the hidden diversity of the monkfish Lophius gastrophysus using integrative taxonomy

The monkfish Lophius gastrophysus is a commercially exploited demersal species distributed along the continental shelf and slope of the western Atlantic Ocean. Like other anglerfish species, L. gastrophysus has a pelagic larval stage that can facilitate dispersal by oceanic currents (Farina et al., 2008). It is considered the most widely distributed species in the genus, spanning more than 10,000 km of the western Atlantic Coast, from North Carolina to Argentina. Previous morphological analyses based on a limited set of traits suggested that L. gastrophysus may comprise two distinct species. We tested this hypothesis by combining molecular and morphometric data. Sequence data were obtained from 38 L. gastrophysus individuals collected from Northern and Southern American populations. We also used a subset of 13 specimens for geometric morphometrics. Phylogenetic analyses and automatic species delimitation methods consistently distinguished Northern and Southern Atlantic L. gastrophysus as separate clades, and geometric morphometrics supported the same split. The congruence among geographic, molecular, and morphometric data indicates that L. gastrophysus harbours two distinct lineages, highlighting the need for a formal taxonomic revision and separate fisheries management of northern and southern Atlantic populations.

genetics↗

Pigmentation markers enable efficient co-CRISPR selection in the pantry moth Plodia interpunctella

CRISPR co-targeting (co-CRISPR) of pigment genes facilitates the identification and recovery of genome-edited individuals in model organisms. Using Plodia interpunctella, an emerging lepidopteran model, we characterized mutant phenotypes of eight genes involved in the melanin and ommochrome biosynthesis pathways (black, ebony, yellow, tan, vermilion, kynurenine 3-monooxygenase/cinnabar (kmo/cn), scarlet, and white) to evaluate their suitability as co-CRISPR markers. Disruption of black, ebony, and yellow altered larval sclerite, pupal cuticle, and adult wing coloration. All four ommochrome pathway genes studied contributed to larval cuticle and adult eye pigmentation, whereas only vermilion and kmo/cn disruption produced visible phenotypes in the adult wings. Sequencing revealed low apparent mosaicism in some yellow and tan G0 individuals, along with recurrent deletions at the yellow and tan target sites consistent with possible microhomology-mediated end joining repair. Using white or scarlet as co-CRISPR selection markers, we identified G0 individuals carrying a secondary mutation in the yellow gene with a positive predictive value of 100% by scoring its knockout phenotype. Extending this approach to six meiosis-related genes without visibly scorable phenotypes, selection based on the white mutant phenotype identified sequence-confirmed mutations at the corresponding gene-of-interest in 191 of 206 G0 individuals (93%). Together, these findings establish white and scarlet as effective co-CRISPR selection markers and provide a practical strategy for facilitating functional genetic studies in Plodia.

genetics↗

Genetic diversity of Plasmodium falciparum and Plasmodium malariae and its association with infectivity of Anopheles coluzzii in the Mfou Health District, Central Region of Cameroon

Background Malaria remains a major public health challenge in Cameroon, with sustained transmission particularly in forest-savannah transition zones. Although Plasmodium falciparum is the predominant malaria parasite, non-falciparum species such as P. malariae also contribute to the human infectious reservoir and to onward transmission. Parasite genetic diversity and multiplicity of infection (MOI) are recognized determinants of malaria transmission and may influence the prevalence and intensity of infection in mosquito vectors. However, whether parasite genetic diversity and MOI differentially influence the infectivity of P. falciparum and P. malariae to local Anopheles coluzzii populations remains poorly characterized. This knowledge gap is particularly relevant in the Mfou Health District, where malaria transmission persists and both P. falciparum and non-falciparum infections occur. Methods A cross-sectional survey was conducted between September 2022 and November 2023 in twelve primary schools in the Mfou Health District. Finger-prick blood samples were collected from children aged 3-17 years and examined by microscopy to identify Plasmodium infections and gametocyte carriers. Gametocyte-positive participants were enrolled in direct membrane feeding assays (DMFA) using locally reared Anopheles coluzzii. Blood samples were tested under two conditions: whole blood (WB) and replacement with naive AB serum. Mosquito infection outcomes were assessed at the oocyst (day 7) and sporozoite stages (day 14). Parasite species were confirmed by quantitative PCR (qPCR), and genetic diversity was assessed using seven microsatellite loci, together with msp1 and msp2 markers for P. falciparum and pmmsp1 for P. malariae. MOI and expected heterozygosity (He) were calculated, and their associations with mosquito infection prevalence and intensity were evaluated using generalized linear models. Results Among 2,807 children screened, 2,111 (75.2%) were infected with Plasmodium, and 388 (18.4%) carried gametocytes. A total of 169 children aged 4-13 years were selected for membrane feeding assays using Anopheles coluzzii, of whom 68% successfully infected mosquitoes. Among 10,872 dissected mosquitoes, 20.2% were infected at the oocyst stage, with both infection prevalence and intensity being higher in mosquitoes fed on AB serum-reconstituted blood than with whole blood. Sporozoite-stage infections were detected in 5.8% of mosquitoes, predominantly under AB serum conditions. Mosquito infection prevalence and intensity increased with gametocyte density, although submicroscopic gametocyte carriers were also infectious. P. falciparum exhibited high genetic diversity (MOI 2.4 {+/-} 0.8; He 0.72 {+/-} 0.07), while P. malariae showed limited diversity (MOI 0.11-0.21; He {approx} 0). Mixed-species and polyclonal infections were associated with enhanced infection outcomes. Moreover, the msp2 FC27 allelic family in P. falciparum and the 800-bp pmmsp1 allele in P. malariae were associated with oocyst development. Overall, parasite genetic diversity, multiplicity of infection, and species composition were positively associated with mosquito infectivity, with only minor seasonal variation observed. Conclusion Malaria transmission in Mfou is sustained by genetically diverse and multiclonal P. falciparum infections, with P. malariae also contributing to the human infectious reservoir. The association of parasite genetic diversity, MOI, and allele-specific factors with mosquito infectivity underscores the importance of addressing asymptomatic and multiclonal infections in malaria control strategies.

genetics↗