torchcell.datasets

Dataset loaders. The top level of torchcell.datasets holds the per-gene embedding datasets (sequence language-model embeddings, codon frequencies, one-hot and random baselines), which derive from the in-memory embedding base in torchcell.data.embedding, and NodeEmbeddingBuilder, which builds them from a configuration. The experiment datasets live in torchcell.datasets.scerevisiae: one module per source publication, each reading that publication’s released data, converting its records into torchcell.datamodels experiments and storing them in LMDB. A class decorated with register_dataset() is added to dataset_registry, the name-to-class map that knowledge_graphs.create_kg and knowledge_graphs.kg_manifest look datasets up in; the table below lists every registered class.

Embedding datasets

NucleotideTransformerDataset	Per-gene Nucleotide Transformer sequence embeddings keyed by window model.
FungalUpDownTransformerDataset	Per-gene embeddings from the fungal up/down-stream sequence transformer.
CodonFrequencyDataset	Embedding dataset of CDS codon-frequency vectors per gene.
OneHotGeneDataset	Embedding dataset producing one-hot vectors over the genome gene set.
ProtT5Dataset	Embedding dataset of ProtT5 protein representations per gene.
GraphEmbeddingDataset	Node-feature embeddings derived from an SGD gene graph.
Esm2Dataset	Embedding dataset that computes ESM2 protein embeddings per gene.
CalmDataset	Embedding dataset that runs the CaLM model over gene CDS sequences.
RandomEmbeddingDataset	Embedding dataset assigning each gene a random feature vector.
NodeEmbeddingBuilder	Builder class for creating node embedding datasets based on configuration.

Dataset registry

register_dataset

Register a dataset class by its name and return it unchanged (decorator).

dataset_registry

Registered S. cerevisiae experiment datasets

The 52 classes in dataset_registry, grouped by source module (one module per publication).

auesukaree2009.EnvChemgenAuesukaree2009Dataset	Auesukaree 2009 six-stress env x geno -> categorical sensitivity dataset.
baryshnikova2010.SmfBaryshnikova2010Dataset	Baryshnikova 2010 genome-scale single-mutant fitness dataset (6023 alleles).
bloom2019.Bloom2019Dataset	Bloom 2019: one record per (segregant, condition), 530,100 records.
cachera2023.BetaxanthinCachera2023Dataset	Genome-wide betaxanthin CRI-SPA product-proxy screen of the YKO collection.
caudal2024.CaudalPanTranscriptome2024Dataset	Natural-isolate pan-transcriptome: each isolate a perturbation set off S288C.
cooper2010.AminoAcidCooper2010Dataset	Cooper 2010 CE-LIF amino-acid pools of the deletion collection (ratio to plate).
costanzo2016.DmfCostanzo2016Dataset	Double-mutant fitness dataset from Costanzo et al. 2016.
costanzo2016.DmiCostanzo2016Dataset	Double-mutant genetic interaction dataset from Costanzo et al. 2016.
costanzo2016.SmfCostanzo2016Dataset	Single-mutant fitness dataset from Costanzo et al. 2016.
costanzo2021.EnvChemgenCostanzo2021Dataset	Costanzo 2021 condition-SGA: env x (deletion | ts-allele) -> differential fitness.
dasilveira2014.MetaboliteDaSilveira2014Dataset	MS lipidome of the yeast kinase/phosphatase deletion collection (127 mutants).
hillenmeyer2008.HetHillenmeyer2008Dataset	FitDb HIP (heterozygous kanMX4 allele replacement) fitness-defect log2-ratio.
hillenmeyer2008.HomHillenmeyer2008Dataset	FitDb HOP (homozygous barcoded kanMX4 deletion) fitness-defect z-score dataset.
hoepfner2014.EnvChemgenHoepfner2014Dataset	Hoepfner 2014 HIP-HOP atlas: env x (het | hom kanMX4 deletion) -> sensitivity score.
kemmeren2014.MicroarrayKemmeren2014Dataset	Deletion-mutant microarray expression profiling from Kemmeren et al. 2014.
kuzmin2018.DmfKuzmin2018Dataset	Double-mutant fitness experiments from Kuzmin 2018.
kuzmin2018.DmiKuzmin2018Dataset	Double-mutant genetic-interaction experiments from Kuzmin 2018.
kuzmin2018.SmfKuzmin2018Dataset	Single-mutant fitness experiments from Kuzmin 2018.
kuzmin2018.TmfKuzmin2018Dataset	Triple-mutant fitness experiments from Kuzmin 2018.
kuzmin2018.TmiKuzmin2018Dataset	Triple-mutant genetic-interaction experiments from Kuzmin 2018.
kuzmin2020.DmfKuzmin2020Dataset	Double-mutant fitness experiments from Kuzmin et al. 2020.
kuzmin2020.DmiKuzmin2020Dataset	Digenic interaction experiments from Kuzmin et al. 2020.
kuzmin2020.SmfKuzmin2020Dataset	Single-mutant fitness experiments from Kuzmin et al. 2020.
kuzmin2020.TmfKuzmin2020Dataset	Triple-mutant fitness experiments from Kuzmin et al. 2020.
kuzmin2020.TmiKuzmin2020Dataset	Trigenic interaction experiments from Kuzmin et al. 2020.
lian2019.CrisprMagicLian2019Dataset	Lian 2019 MAGIC per-guide CRISPRa/i/d furfural-enrichment env x geno dataset.
lopez2024.IsobutanolScreenLopez2024Dataset	Montaño López 2024 FIRST genome-wide isobutanol-biosensor YKO screen (Table S2, n=1).
lopez2024.IsobutanolValidatedLopez2024Dataset	Montaño López 2024 VALIDATED isobutanol-biosensor re-screen (Table S3, triplicate).
messner2023.ProteomeMessner2023Dataset	Microflow-SWATH-MS proteome of the yeast gene-KO collection (4,699 strains).
mormino2022.CrispriMormino2022Dataset	Mormino 2022 CRISPRi acetic-acid biosensor screen (12 isolated strains).
mota2024.EnvChemgenMota2024Dataset	Acetic/butyric/octanoic acid chemogenomic env x geno -> ordinal susceptibility.
mulleder2016.AminoAcidMulleder2016Dataset	Genome-wide intracellular amino-acid metabolome of the yeast deletion collection.
nadal_ribelles2025.NadalRibellesPerturbSeq2025Dataset	Nadal-Ribelles 2025 pseudobulk Perturb-seq: one record per (genotype, condition).
oduibhir2014.SmfODuibhir2014Dataset	O'Duibhir 2014 single-deletion relative-growth-rate -> fitness dataset.
ohnuki2018.ScmdOhnuki2018Dataset	CalMorph morphology of essential-gene heterozygous diploid deletions (Ohnuki 2018).
ohnuki2022.ScmdOhnuki2022Dataset	CalMorph morphology for 1979 quadruple (target + 3Delta) deletion strains.
ohya2005.ScmdOhya2005Dataset	CalMorph morphology for haploid non-essential single-gene deletions (Ohya 2005).
ozaydin2013.CarotenoidOzaydin2013Dataset	Beta-carotene colony-color visual screen of the yeast deletion collection.
sameith2015.DmMicroarraySameith2015Dataset	Double mutant GSTF expression profiling from Sameith et al. 2015.
sameith2015.SmMicroarraySameith2015Dataset	Single mutant GSTF expression profiling from Sameith et al. 2015.
sgd.GeneEssentialitySgdDataset	Gene essentiality experiments built from SGD inviable null phenotypes.
smith2006.FattyAcidSmith2006Dataset	Smith 2006 fatty-acid clear-zone env x geno -> ordinal response dataset.
smith2016.CrispriChemgenSmith2016Dataset	Smith 2016 per-guide dCas9-Mxi1 CRISPRi chemical-genetic env x geno dataset.
synth_leth_db.SynthLethalityYeastSynthLethDbDataset	Yeast synthetic lethality gene-pair experiments from SynLethDB.
synth_leth_db.SynthRescueYeastSynthLethDbDataset	Yeast synthetic rescue gene-pair experiments from SynLethDB.
vanacloig2022.EnvChemgenVanacloig2022Dataset	Anaerobic chemical-genomic env x geno -> log2(inhibitor/control) response screen.
wildenhain2015.EnvChemgenWildenhain2015Dataset	Wildenhain 2015 chemical-genetic matrix: env x geno -> growth-inhibition z-score.
xue2025.FattyAcidXue2025Dataset	Xue 2025 in-house FFA combinatorial-deletion titer dataset (5 FFAs/strain).
yeastphenome.YeastPhenomeDataset	Curated YeastPhenome growth-phenome env x geno -> NPV z-score screens.
yoshida2012.OrganicAcidYoshida2012Dataset	HPLC organic-acid titers of 17 yeast deletion mutants (Yoshida 2012, Table 3).
zelezniak2018.MetaboliteZelezniak2018Dataset	LC-SRM targeted metabolome of the kinase-knockout collection, one record per (strain, protocol); each record and its WT reference share one protocol.
zelezniak2018.ProteomeZelezniak2018Dataset	SWATH-MS proteome of the yeast kinase-knockout collection (97 strains).