What We Do – BioLizard

Multi-omics data analysis

Biomarker detection & validation

The identification of biomarkers for early disease screening and therapy monitoring can be enhanced through the use of predictive algorithms using single or multi-omics data. Our extensive experience in the development of artificial intelligence based algorithms supports the de novo discovery and validation of biomarkers using unique proprietary clinical diagnostic algorithms. Next to making use of single or multi-omics approaches, we are also proficient in the computational integration of additional clinical and pathological data using multi-modality methods.

Skills

Drug candidate discovery

Protein alteration is prevalent in disease therefore they may be more or less active, have an altered function. Small molecule drugs may exist that modify a protein’s function by activating, inhibiting or altering the protein. We generate structural models of the disease altered proteins of interest, implementing a high throughput in silico workflow to screen multiple databases and score relevance of small molecules as clinical drug candidates.

Skills

Epigenome analysis

Both global and local DNA methylation changes have been associated to disease, and in particular to cancer. While methylation is the most well studied for of epigenetic modification, other non-methylation epigenetic mechanisms include histone modifications, micro-RNA interactions and chromatin remodelling complexes. We build robust computational pipelines for full epigenetic analysis to detect differences in methylation and non-methylation mechanisms for selected treatments and link these results to gene expression.

Skills

Microbiome analysis

We perform multi-omics mapping of full microbial communities to determine species composition and examine the causal role of a specific microbiome in selected disease states. Our workflows are used for human, animal, plant and aquatic systems. We implement state-of-the-art statistics to calculate causal statistics by comparative weighting of genetic, microbial and external factors to predict disease development, progression and resistance. By using a combination of a biostatistical frameworks and artificial intelligence tools we examine modifiable risk factors to establish the most viable drug targets and biomarkers in microbiome-associated illnesses as well as develop bioremediation approaches.

Skills

Protein profiling & proteomics

Our proteomic data analysis workflow uses mass-spectrometry data in order to identify and quantify peptides and proteins. Our pipelines can analyse peptides and their post-translational modification as well as full proteomes. Integration of transcriptomic and proteomic data highlights correlation of specific proteins and transcripts. Using proteogenomics analysis (integration of genomic and transcriptomic data) we can build a custom protein databases.

Skills

Structural variant calling

Structural variants, for example fusion genes formed as a result of translocation, interstitial deletion, or chromosomal inversion have been linked to cancer development. Using targeted RNA-sequencing data our pipeline detects clinical fusion genes and by further implementing predictive algorithms we can predict cancer prognosis, patient survival and treatment responses. Additionally we can integrate publicly available drug data to screen drug candidates for potential therapeutic treatments for fusion gene inhibition.

Skills

Transcriptomic data analysis

Exploring RNA-sequence data using tailor made pipelines that are reproducible and robust to accurately identify transcriptional differences. Transcriptomes are investigated using a combination of in-house developed and publicly available tools to effectively assemble and align RNA-sequence reads. This can be done for samples with or without a reference genome. Differential expression is determined at a gene or transcript level and potential pathways are highlighted through gene enrichment analysis. Our pipeline, toRNAdo, is a Nextflow based pipeline, that allows a fast scaling processing on different cloud technology (Google cloud , Amazon AWS) to RNA seq from fastq files up to obtain read count per gene. In a fully automated fashion, toRNAdo provides both final result and also QC report based on MultiQC tool. Due to its modularity it is quite easy to modify in order to process other kind of biological data such as variant calling.

Skills

Algorithm development

Biomarker detection & validation

Skills

Clinical data trials

Repurposing historical data and existing real-world data to build effective models able to predict clinical outcomes. We use federated learning techniques to develop artificial intelligence based trial design, improve patient enrolment, predict retention and drive artificial intelligence enabled clinical trial analytics. We can also use real-world evidence to build synthetic control arms for clinical trials.

Skills

Epitope prediction

Optimise antigen prediction platforms in the area of specific oncology indications by developing artificial intelligence based bioinformatics algorithms to predict binding sites and binding efficiency. On top of this, a reporting element is incorporated into the platform developed presenting the algorithm output in a clear and concise report interpretable by medical personnel.

Skills

Image detection, recognition & classification

Using a combination of machine learning, deep learning and imaging data informatics we can enhance image processing and accuracy for a complete range of image driven solutions. Integrating multi-omics and histological data with imaging data we can predict tumour margins, patient prognosis, most minimally invasive treatments and therapeutic responses on a case by case basis. Artificial intelligence based algorithms can be integrated into medical devices providing real-time data analysis and an artificial intelligence aided interpretation system.

Skills

Multimodal learning

We use multiple omics data sources to provide a holistic approach to data learning by developing learning models that effectively integrate different data features simultaneously. By doing so, we can discover trends hidden in multiple data sources that only become evident when they are included together in the learning process. Switching to multimodal predictions also enables more robust predictions than a unimodal learning system.

Skills

Personalised medicine

Personalised medicine or precision medicine makes use of an individual’s clinical, genetic, genomic, and environmental information. Using a combination of data types (multimodal testing), we develop and optimise robust models calculating patient risk score for specific diseases (eg. cancer), treatment dosing and responses, survivability as well as identifying predictive biomarkers for disease.

Skills

Data mining & integration

Federated learning

Balancing the predictive performance of an algorithm by feeding it more data with the concerns for data privacy is an important issue. Federated Learning solves this by bringing the algorithm itself to the data. The data never leaves the end user at all time, on-device training is performed and only the updates to the algorithm are reported back to the central server.

Skills

Image detection, recognition & classification

Skills

Multimodal learning

Skills

Personalised medicine

Skills

Clinical data integration and management

Capturing, integrating and storing data from clinical trials, patient reported outcomes and wearables into a database accessible through an easy to use platform. By combining artificial intelligence algorithms and wearable technology we develop easy to use apps to automate data capture and continuously monitor and manage patients.

Skills

Pipeline & app development

Clinical data trials

Skills

Epitope prediction

Skills

Federated learning

Skills

Personalised medicine

Skills

Clinical data integration and management

Skills

Computational pipeline integration

Our experts facilitate the development, deployment and follow up of complete bioinformatics pipeline frameworks integrated into specific platforms or devices. These can be supported by web-based applications that can not only trigger the execution of a pipeline, but also include a reporting component that renders the output in a fully customisable format that can be interpreted by non-technical users.

Skills