ML6 for lifesciences

Facilitating progress in lifesciences with AI

By applying our broad AI knowledge stack on biology, ML6 aims to make a positive impact on the world by accelerating drug discovery, enzyme development, and biological data analysis.
‍

Challenges Solutions Cases Contact us

The Future is Autonomous

Explore how AI Agents are revolutionizing industries! Learn more about the groundbreaking advancements and why they're a game-changer for your business.

Find out more

Challenges

Lifesciences are essential in an ever changing environment. The intricate study of biological molecules, genes, and cellular processes has revolutionised our understanding of life itself, offering groundbreaking insights into health, genetics, and disease.
Artificial intelligence, on the other hand, has made huge technological improvements in the last decade. AI empowers researchers parsing large datasets, revealing undiscovered patterns and or different angles to work with data, e.g. the protein folding challenge

ML6 is here to help along each step of the way.

Multi-omics analysis

Analyzing diverse data from genomics, transcriptomics, proteomics, and metabolomics poses challenges due to complexity and noise. Machine learning effectively identifies patterns in these datasets, providing valuable insights across various research areas.

Protein understanding

Proteins drive biochemical processes and have wide-ranging applications from drug development to sustainable practices. Machine learning models like Alphafold3 and ESMFold decode protein language, but challenges remain, such as protein design. Implementing these models into daily workflows requires robust data pipelines, where MLOps can streamline lab processes.

Genotyping-Phenotyping

Standardizing phenotypic data collection is crucial for pesticide development and breeding, with machine learning aiding in standardization. Standardized phenotyping facilitates linking genotype to phenotype, crucial in genetic breeding and cross-lab comparisons.

ML6 expertise

Expert AI solutions for the lifesciences industry

At ML6, we understand the unique challenges faced by the lifesciences industry. Our extensive knowledge and AI expertise enable us to develop practical solutions tailored to your business needs. Here's how we help lifescience companies tackle their specific challenges:

Drug development

We help building structure function relationships, using machine learning. Machine learning lends itself very well to capture as much information as possible to help steer the drug development process.

Examples

Biomarker identification

Drug repurposing

Optimisation of clinical trials

Molecular diagnostics

Molecular diagnostics, a thorough analysis of genomics and proteome, yields important insights on how diseases work, where risks lay, and also what therapies might or might not work. AI is the tool to unravel complex patterns in these complex datasets, finding earlier warning signals in disease development, picking up on new biomarkers,...

Examples

Disease prediction & risk assessment

Biomarker discovery

Predictive analytics for disease progression

Multi-omics analysis

Based on the information obtained through forecasting and simulation, both the design and the operation of systems can be optimised. In the design phase, optimization can be applied to network topology, wind farm layout, and EMS configuration. During operation, optimization can improve congestion management, wind turbine steering, and EMS control.

Examples

Pattern recognition & data mining

Integration of multi-omics data

Predictive modeling for disease diagnosis & treatment

Biotech acceleration

Research in machine learning with life science applications, e.g. Alphafold2 and ESMFold, have demonstrated how it can accelerate life science and biotechnology.Automated pipelines for e.g. protein engineering process with the latest models and data processing blocks, help researchers to maximise experimental value, and shorten time to market. Shortening development time can mean significant cost savings in the entire development process.

Examples

Protein structure predicition

Automated pipeline development

Proven Expertise

Success Stories

Plant phenotyping - a more data driven approach from lab to field

Registering observable features of plants, plant phenotyping, forms an essential part of (genetic) plant breeding, fertiliser development, pesticide development,... A good and objective phenotyping pipeline can serve as a direct driver for further research.
ML6 has developed such a standardised phenotyping pipeline for objective scoring of plants for the development of novel pesticides.
Albeit, it is important to understand how these products work in the lab, putting it to use in the real world is equally important. Hence the data collection tooling for the pipeline was developed in an app for mobile devices. This enabled data collection from the first lab experiment until on the field. The information returned is therefore not only interesting for feedback on specific products, but can also be used as a benchmark for the used product discovery pipeline.By highly automating the computer vision observation pipeline to score the health of plants as well as the broader consistent application of the tool in the lab up to the field, the product development line has become more data driven, which in turn shortens the time to market as well as improving the usability of the products.

Drug repurposing - leveraging extensive knwoledge graphs

Drug development is a very resource intensive process. The development phase as well as the thorough testing, takes time and can be costly. On the other hand, it has been shown that many drugs have beneficial side effects. E.g. Aspirin was originally developed as a painkiller, but it is now repurposed to treat cardiovascular diseases and colorectal and breast cancer.
Drug repurposing has the advantage that the drug already exists, the production process is in place and that the regulatory tract can be shortened. By using an extensive knowledge graph as a basis, and the addition of extra datasets, we were able to predict a shortlist of drugs that would be worth testing for the blocking of HIV.
Different algorithms were used, based on the drug - gene - disease links, over new link predictions, but also molecular structure.Some of drugs in the short list were confirmed to have a HIV blocking activity in the literature, without this knowledge being in the original dataset.Biological processes are intrinsically complex, and cover different processes, which very often lead to diverse datasets. Insights can be generated from these individual datasets, however the true power comes in combining different datasets, generating more in depth insights on the processes.

deCYPher - minimising lab experiment whilst maximising experimental value

The use of microbial cell factories has the potential to accelerate the sustainable production of common chemicals, as well as it opens up the possibility to produce high added value molecules.ML6 participates in deCYPher, a Horizon Europe project. The aim of the project is to produce high added value terpenoids and flavonoids for which current production processes are limited. The project will cover the complete development cycle, starting from protein discovery and engineering, over microbial cell factory design and setup, over to small scale production runs.
ML6 will collaborate with the deCYPher partners, speeding up development cycles. Furthermore, we want to guide the biotech field by transferring our board knowledge from other industries into the biotech field. This covers computer vision, natural language processing and structured data applications.
Lab experiments are almost always hard, and resource costly: time, materials and machines. For deCYPher ML6 will support the wetlabs with machine learning pipelines that will suggest experiments with maximal value, for learning as well as for the end goal.

Protein heat stability - protein folding

Biocatalysis is a powerful way of producing high end molecules. The process is simple, fast and precise, as it uses a limited amount of enzymes, products and substrates.
An economical competitive process is a process that can run for prolonged time. Improved (heat) stable enzymes ensure that the process can be run for longer periods, improving on the economical viability.ML6 participated in a Kaggle competition for the prediction of protein heat stability. Novozymes supplied an initial dataset, and the goal was to predict heat stability of different protein variants.
We’ve built a data pipeline with extensive data preprocessing and reformatting, and that uses the strengths of models like Alphafold2 and ESMfold to make a global prediction.Proteins are at the core of life, as well as of life sciences. Machine learning has given an enormous push to protein understanding. Protein folding can be a crucial answer for question in very diverse fields. At ML6 work and use the latest protein models in useable pipelines in order to bring the value of these models to the end user.