> ## Documentation Index
> Fetch the complete documentation index at: https://instance.bio/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# Release Notes

> Dataset and pipeline updates.

## 26.4 (2026-09-15)

### Dataset improvements

* `expression_scores` has been retired in favor of consolidated `binding_scores`.
* The `binding_scores` and `replicate_binding_scores` have been improved to account for expression and dropouts. These dataframes now account for every target-candidate pair even if there was no binding score recorded. The new `is_expressed:bool` column tracks if the candidate binder was successfully expressed and `expression_umi_count:number` reports the UMI diversity of the input into the binding process.

### Process improvements

* Multiple optimizations to the binding process parameters to increase UMI recovery.

## 26.3 (2026-06-30)

This release focuses on optimizing the antigen expression process, binding process, and sequencing analysis to improve the ability to characterize weaker binders with greater sensitivity.

### Dataset improvements

* Data labeling task metadata and UMI aligned-read provenance datasets are included alongside the primary datasets:
  * `dim_candidates`
  * `dim_samples`
  * `aligned_reads`
* Binding scores now take bind and wash process noise into account using negative control samples.
* UMI clustering was updated to improve unique molecule counts.
* Aligned-read filtering now includes only perfect coding regions.

### Process improvements

* Binding incubation parameters were adjusted to allow weaker binders to reach equilibrium more consistently.
* A flash-wash protocol reduces bead binding time and end-to-end wash time.
* Default antigen-dose parameters improve sensitivity across a broader range of candidate binding strengths.
* Negative controls now include a null-antigen sample of expressed candidates that undergoes the identical binding process.
* Binder expression process conditions were optimized to improve yield uniformity.

## 26.2 (2026-06-12)

This minor update addressed usability issues with exposed data and schemas.

* Internal replicate details, including UMI counts, are now available in standard datasets:
  * `expression_replicates`
  * `binding_replicates`
  * `specificity_replicates`

## 26.1 (2026-05-29)

The initial production-ready release of Instance Data Lab.

* Initial production release of the protein experimental ground-truth generation harness.
* Baselined and benchmarked against open-source datasets.
* Support for SinoBiological recombinant protein soluble antigens.
* Support for candidate binders up to 175 amino acid residues, including mini-protein and VHH designs.
* Support for `expression`, `binding`, and `specificity` data labels.
* Native support for replicates, multiple antigen doses, and unique molecular identifiers (UMIs).


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.