🪵 Latch Changelog — October 2024

<aside> 💡

Highlight of the month: We released a comprehensive suite of 20+ protein engineering tools. With these, we constructed an example virtual drug screening pipeline to create de-Novo binders for PCSK9—a protein involved in cardiovascular disease and of active interest in several modern drug programs

</aside>

https://twitter.com/kenbwork/status/1847298319653142726

<aside> 💡

Our engineers quietly rolled out 40+ bug fixes and stability improvements across the platform, making Latch smoother and more reliable for you.

</aside>

• Data

  • Enabled navigation to locations in LData from past execution Inputs pages.

    navigate-to-ldata.mp4

  • Displayed S3 links for file paths in Registry.

    egenesis s3 linking.mp4

  • Added automatic retry by reducing chunk size for latch cp to enable more stable data uploads in poor Internet connection conditions.

• Registry
  • Improved loading performance for large Registry tables with 300K+ rows
• Workflows
  • Nextflow Integration

    • Introduced latch develop command to attach OFS volume to a pod for inspection of intermediate files.

    • Every Nextflow execution now comes with a per-process usage report, giving developers higher visibility into resource consumption for pipeline optimization.

      usage-report.mp4

    • Enabled relaunching successful executions with the existing workdir.

    • Displayed .command.sh, .command.out, and .command.err files for each process.

    • Improved stability of OFS storage by addressing common issues: files not being available on all mounts immediately and unaccounted memory of OFS pod. We added memory over-provisioning to workflow tasks, shipped a custom fsync binary and made the OFS runtime process more stable which mitigated transport disconnect errors in the workflows.

    • Released bug fixes and stability improvements:

• Plots
  • PlotsAI

    • For Plots’ built-in LLMs, our engineers optimized prompts to guide the LLMs in using specific APIs and assay-specific guides, ensuring responses are grounded in reliable tools for answering complex biological questions.

      

      The result:

      msigdb.mp4

    • Introduced o1-preview option in LLM settings

  • Improvements to Plots’ DX:
    • Enabled interruption/killing of running cells
    • Ranked options by edit distance in w_select and w_multi_select.
    • Moved code cell to the center view
    • Added full screen mode for every code cell
    • Added rendering matplotlib and seaborn figures in code cell output
    • Added subsampling for plots with 50K+ points, improving loading performance of large plots in the browser
    • Enabled per-cell widget value clearing.
    • Reduced connection times from average of 45 seconds to 25 seconds
    • Added cell tracking during execution with scroll-to functionality
    • Added sorting by latest opened on the Plots notebooks’ organization view
  • Bug fixes and stability improvements:
• Kit Provider
  • Released bug fixes and improvements

🪵 Latch Changelog — September 2024

• Platform:

  • White-labeled analysis portal for kit providers:

    <aside> 🎯

    We launched a new product for solutions and kit providers, offering their customers access to a white-labeled analysis portal to their customers 🎉

    Every customer can redeem a code an access a workspace custom to them with their raw data, a no-code GUI for launching custom workflows, and Plot templates for interactive tertiary analyses.

    Kit providers can fully configure all components of each customer workspace, allowing them to control and tailor the optimal experience for their customers.

    </aside>

    • The customer journey looks like the following:
      • They purchase a kit.
      • Scan a QR code and/or receive an email with a link that invites them to their custom workspace.
      • Once the customer clicks on the invitation link, they get taken to a beautiful, white-labeled portal for that kit provider
      • The customer logs in, redeems their code, and receives a number of credits for analysis (As a provider, you can customize this number in your Admin panel)
      • They see their raw data, workflows, and plot templates that the provider has specified for them (All these components are also customizable by the kit provider in their Admin panel)

• Data:

  • latch cp now works natively on Windows, which eases the process of uploading data from Windows computers attached to instruments to Latch.

• Registry:

  • Supports bulk import of sequencing files into a table in Registry.
    • By clicking on a folder that contains FastQ files, Latch’s importer will intelligently detect forward & reverse reads, as well as the lane to which they belong, and properly populate them into a sample sheet format in Registry. This minimizes mistakes from scientists filling out sample sheets that could cause expensive workflow failures downstream.

• Workflows:

  • Our engineers verified and released no-code interfaces for the top 10 nf-core pipelines on Latch, using our native Nextflow integration.
  • Read more about our recent Nextflow announcement here.

  

• Nextflow:

  • Released ObjectiveFS as the new shared filesystem underlying all Nextflow processes
  • This gives developers the flexibility of choosing another shared filesystem that is cheaper than EFS.

• Plots:

  • Released new pricing for Latch Plots
    • Please see our full pricing announcement below.
  • Better developer experience: Implemented Language Server Protocol (LSP) into Latch Plots to enable better developer experience with syntax highlighting, display function definition upon hovering
  • Shareable links: Scientists can now capture a snapshot of their Plot notebooks, including all input parameters and plots, and generate a read-only link. This feature ensures the immutability of scientific records and analyses, enforcing reproducibility best practices. Such practices set teams up for success in future IND filings.