seqoutlm-4B-GGUF

SeqoutLM 4B

SeqoutLM is a specialized biomedical metadata normalization model that converts unstructured genomic sample metadata into a standardized JSON representation consisting of 16 fixed fields.

The model is designed for large-scale metadata harmonization across public genomics repositories such as GEO and SRA, enabling downstream search, filtering, integration, and analytics workflows.

Output Schema

The model always returns a JSON object containing the following fields:

• organism
• tissue
• tissue_primary_site
• tissue_site_type
• cell_type
• cell_line
• disease
• phenotype
• strain
• ethnicity
• development_stage
• treatment
• genetic_modification
• assay
• assay_category
• sample_type

All fields are always present. When a value cannot be determined from the input, the model outputs null.

Example

Input

{
   "title":"Single cell transcriptomics reveals conserved regulatory networks in human and mouse interneuron development",
   "summary":"Inhibitory GABAergic interneurons originate in the embryonic medial ganglionic eminence MGE and control network activity in the neocortex Dysfunction of these cells is believed to lead to runaway excitation underlying seizure based neurological disorders such as epilepsy autism and schizophrenia Despite their importance in heath and disease our knowledge about the developmentof this diverse neuronal population remains incomplete Here we conducted single cell RNA sequencing scRNA seq of human fetal MGE from 10 to 15 weeks post conception TheseMGE tissues are composed of largely cycling progenitors and immature post mitotic interneurons with characteristic regional marker expression Analysis of integrated humanand mouse MGE data revealed species conserved transcriptomic profiles and regulatory program Moreover we identified novel candidate transcription regulators for human interneuron differentiation These findings provide a framework for in vitro modelling of interneuron development and strategy for potentially enhance interneurons production from human pluripotent stem cells",
   "overall_design":"Time course single cell RNAseq from the human embryonic medial ganglionic eminence MGE from post conceptional weeks 10 13 14 and 15",
   "attributes":{
      "tissue":"medial ganglionic eminence (MGE)",
      "group":"A_13pcw",
      "developmental stage_pcw":"13",
      "batch":"batch2",
      "sample id":"2532",
      "source":"medial ganglionic eminence (MGE)",
      "molecule":"polyA RNA",
      "sample_title":"MGE_A_13pcw_709",
      "sample_description":"MGE_A_13pcw_709"
   }
}

Output

{
   "ethnicity":null,
   "phenotype":null,
   "cell_type":"Interneuron",
   "tissue":"Medial Ganglionic Eminence (Mge)",
   "strain":null,
   "disease":"Healthy",
   "assay":"scRNA-Seq",
   "assay_category":null,
   "cell_line":null,
   "treatment":"Mge_13Pcw",
   "development_stage":"Embryonic",
   "sample_type":"Primary Tissue",
   "genetic_modification":null,
   "organism":null,
   "tissue_primary_site":null,
   "tissue_site_type":null
}

Acknowledgements

This model was trained using the Unsloth framework on top of Qwen 2.5 0.5B Instruct and the Seqout metadata normalization dataset.

<img src="https://raw.githubusercontent.com/unslothai/unsloth/main/images/unsloth%20made%20with%20love.png" width="200">