Contract Development Services by Lumetics

Lumetics presenting on Machine Learning at AAPS San Diego • May 12 • 2:30PM

Contract Development Services

Particle Classification

Prepare high-quality training datasets from control samples with ease, and seamlessly apply your proprietary machine learning models for automated particle classification—directly within the Lumetics LINK™ platform.

Model Development

Lumetics develops and optimizes TensorFlow-based prediction models using your training and validation datasets. All resulting models are fully owned by the end user and can be seamlessly deployed within the LINK platform.

Deployment

Lumetics deploys and validates machine learning prediction models on a high-performance TensorFlow Server within your local network, enabling reliable, routine application directly from within the LINK platform.

Machine Learning Example

This example demonstrates the integration of a TensorFlow-based machine learning workflow within the LINK platform. A supervised learning model is trained and validated using user-provided datasets, then serialized and deployed into a local TensorFlow Serving instance.

Within LINK, the model is executed as part of standard analytical workflows, enabling deterministic, reproducible predictions directly from imported data. The deployment supports both batch and routine in-line inference, with results returned as structured outputs for downstream analysis and reporting.

This implementation illustrates LINK’s capability to operationalize machine learning models in an on-premise environment while maintaining controlled data flow, versioned model deployment, and integration with existing analytical pipelines.

S-Factor Optimization

S-Factor Development

S-Factor Optimization is a data-driven service within the LINK platform that systematically tunes particle classification and detection parameters to maximize analytical accuracy and consistency. Using real instrument data, LINK evaluates and optimizes the S-factor—an internal scaling parameter that governs particle detection sensitivity and classification thresholds—ensuring optimal separation of particle populations (e.g., protein vs. silicone oil) across defined size ranges.

Functional Overview

Calibrates detection sensitivity to balance signal vs. noise
Optimizes particle classification across size bins (e.g., 2–10 µm, ≥10 µm, ≥25 µm)
Reduces false positives/negatives in sub-visible particle analysis
Improves reproducibility (lower variability across measurements)
Aligns results with regulatory expectations (e.g., USP <788>, <789>)

LINK Delivery

LINK ingests raw instrument data and applies iterative optimization across S-factor ranges to refine detection performance. Results are benchmarked against known or expected distributions to ensure accuracy and consistency. Optimized parameters are then validated within a controlled, audit-trailed environment before being deployed directly into production workflows.

LINK provides pre-configured dashboards to support S-Factor optimization visualization within the platform.

Outcome

The service delivers more accurate particle counts and classification, increasing confidence in protein versus non-protein differentiation. It also reduces method development time while ensuring consistent, inspection-ready results across analyses.

S-Factor Optimization Example: MFI Analysis with LINK

This example highlights how LINK applies S-factor optimization to improve particle classification in Micro-Flow Imaging (MFI) data, enabling accurate differentiation between protein aggregates and silicone oil droplets. By analyzing morphological parameters and optimizing S-factor cutoffs across particle size ranges, LINK significantly reduces misclassification and improves consistency.

The result is more accurate, reproducible particle characterization with faster method development and seamless deployment into production workflows—delivering inspection-ready, regulatory-aligned results.

More contract development offerings

Software/Database Development

Lumetics specializes in custom software development to extend LINK and support your unique workflows. From machine learning models and advanced data processing to tailored database solutions and feature enhancements, we build exactly what your lab needs.

Morphological Filter Pameters

Software/Database Development

Lumetics provides custom software development to create and optimize TensorFlow-based prediction models using your training and validation datasets. All resulting models are fully owned by the end user and can be seamlessly deployed within the LINK platform, ensuring efficient data processing and effective machine learning integration.

Instrument Support

Software/Database Development

Instrument Support

Lumetics specializes in custom software development by deploying and validating machine learning prediction models on a high-performance TensorFlow Server within your local network. This approach enables reliable data processing and routine application directly from within the LINK platform.

Contract Development