Home
Artificial intelligence and machine learning for health, and
beyond!
About us
Pharmatics overview:
✦ We accelerate intelligent data analysis. We are
experts in machine learning, data sciences, and bioinformatics.
We offer consultancy services in data science and data mining, and we
develop intelligent
tools and analytics engines tailored to your tasks. Where relevant,
we co-develop with your team,
or work closely with recognized experts in your application area.
✦ We use Artificial Intelligence and Machine Learning to
develop a portfolio of intellectual
property
in multiple areas of precision medicine and health analytics. We work
in tandem with internationally recognized
experts in computational genetics, glycomics, and
population health. We partner with opinion leaders
in telemedicine and telehealth, COPD, diabetes,
colon cancer, head and neck cancer,
and rheumatoid arthritis. We work closely with former NHS
executives on improving efficiency of health services.
What we do:
✦ Data Analytics. We help to convert data to
actionable insights for your organization.
We provide consultancy services on converting your data into actions,
products, and decisions.
✦ Machine Learning and AI Engines. We help to convert your services, products, and mobile apps into intelligent AI-powered systems. We can power your solutions with intelligent algorithms to maximize the value for you and your customers.
✦ Privacy-preserving predictive analytics. When individual-level patient data is limited or cannot be easily shared, we may still be able to provide personalized predictions based on advanced ways of combining a priori probabilities with population-level summary statistics.
✦ Machine Learning and AI Engines. We help to convert your services, products, and mobile apps into intelligent AI-powered systems. We can power your solutions with intelligent algorithms to maximize the value for you and your customers.
✦ Privacy-preserving predictive analytics. When individual-level patient data is limited or cannot be easily shared, we may still be able to provide personalized predictions based on advanced ways of combining a priori probabilities with population-level summary statistics.
✦ Precision Medicine. We help to improve
efficiency of drug discovery and drug development
by identifying which interventions work, for whom, and under which
conditions. We can enrich your
studies with our annotations, risk scores, or pre-trained models
developed for a number of indications.
We offer access to our growing IP
portfolo in precision medicine.
✦ Data Cleansing. We are experts in data enrichment, imputation, integration, and harmonization. We develop custom-made tools tailored to molecular acquisition platforms and clinical records. We offer cutting-edge, in-house, solutions for liquid chromatography and next-generation sequencing.
✦ Health intelligence. We help to improve effectiveness of healthcare via advanced data analytics. We provide data-driven interactive summaries, in-depth reports, or implementation plans on how to achieve cost and performance targets.
✦ Data Cleansing. We are experts in data enrichment, imputation, integration, and harmonization. We develop custom-made tools tailored to molecular acquisition platforms and clinical records. We offer cutting-edge, in-house, solutions for liquid chromatography and next-generation sequencing.
✦ Health intelligence. We help to improve effectiveness of healthcare via advanced data analytics. We provide data-driven interactive summaries, in-depth reports, or implementation plans on how to achieve cost and performance targets.
Pharmatics research:
✦ Lead SME in FP7 Statistical Methods for Analysis
of -omics Data
✦ Lead SME in FP7 Machine Learning for Personalized
Medicine
✦ Several ongoing international and UK collaborations
on specific indications and smart healthcare
Pharmatics IP portfolio
Pharmatics expertise
We have in-house expertise in:
✦ High-dimensional classification, regression, survival
analysis
✦ Approximate inference and learning
✦ Representation learning and deep learning
✦ Approximate inference and learning
✦ Representation learning and deep learning
✦ Digital signal processing
✦ High-performance computing
✦ Machine learning and artificial intelligence
✦ High-performance computing
✦ Machine learning and artificial intelligence
✦ Clinical and molecular epidemiology
✦ Population health
✦ Causality
✦ Population health
✦ Causality
✦ Statistical genetics and bioinformatics
✦ Biomarker discovery
✦ Risk stratification
✦ Biomarker discovery
✦ Risk stratification
We handle heterogeneous types of data including:
✦ Clinical records
✦ Telemetry and sensors
✦ Activity and symptoms
✦ Telemetry and sensors
✦ Activity and symptoms
✦ Summary statistics
✦ Annotation databases
✦ Web aggregators
✦ Annotation databases
✦ Web aggregators
✦ Genomics, transcriptomics
✦ Metabolomics, lipidomics
✦ Glycomics, proteins, immunoassays
✦ Metabolomics, lipidomics
✦ Glycomics, proteins, immunoassays
✦ Payer and provider data
✦ Prescription data
✦ Health quality indicators
✦ Prescription data
✦ Health quality indicators
We have established partnerships with internationally recognized experts in:
✦ Computational genetics
✦ Glycomics and glycoproteomics
✦ Glycomics and glycoproteomics
✦ Telemedicine and telehealth
✦ Health analytics
✦ Health analytics
✦ Chronic obstructive pulmonary disease (COPD)
✦ Diabetes and diabetic complications
✦ Diabetes and diabetic complications
✦ Colon cancer and oral cancer
✦ Common autoimmune diseases
✦ Common autoimmune diseases
Methods
- We use our in-house platform for rapid training, evaluation, and comparison of hundreds of machine learning methods. This either solves your problem or guides our development of a custom-made model tailored to your task.
- Our methods are general-purpose. However, they are especially suited for the settings where data is expensive to collect or share, or where the outcomes of interest are rare.
- We maximize the value of data. We rescue unsuccessful studies by finding a better use for your data. We use transfer learning and representation learning to train models with x10-x100 fewer samples.
- Our special expertise is machine learning and AI for drug development, medicine, and health analytics.
✦ Data cleaning, harmonization, integration. This includes
automatic extraction
of useful signals from raw data, outlier detection and correction,
linkage and harmonization,
correction for batch effects and possible confounders, and
inference of missing measurements.
✦ Data enrichment. We use relevant annotation databases, risk scores, and feature libraries to inform our models. Where relevant, we use information from published scientific articles or web-based information aggregators, or our own pre-trained models.
✦ Predictive analytics for classification, regression, and survival analysis. We balance accuracy and model complexity. To maximize the value of data, we use regularized methods, similarity-based methods, deep learning, and ensembles bundled with data-enrichment and feature-learning techniques, among others.
✦ Data enrichment. We use relevant annotation databases, risk scores, and feature libraries to inform our models. Where relevant, we use information from published scientific articles or web-based information aggregators, or our own pre-trained models.
✦ Predictive analytics for classification, regression, and survival analysis. We balance accuracy and model complexity. To maximize the value of data, we use regularized methods, similarity-based methods, deep learning, and ensembles bundled with data-enrichment and feature-learning techniques, among others.
✦ Evaluation via stress-testing: resampling techniques,
external datasets, or vignettes
tailored to your task. Our techniques account for the possible
heterogeneity of covariates and outcomes, differences between training
and validation
datasets, class imbalances, and cost asymmetries.
✦ Learning from summaries. For some tasks, data is expensive to share between multiple sites. We can combine prior knowledge with summary statistics to get close to the state-of-the-art, without accessing individual-level data. This can be used for model initialization and also for meta-analysis, prediction, and causality.
✦ Multiple data types. At present we can handle high-dimensional stationary data and longitudinal (time series) data including health records, sensors, telemetry, user-generated data, activity data, web information aggregators, genomics, transcriptomics, metabolomics, proteomics, and glycomics.
✦ Learning from summaries. For some tasks, data is expensive to share between multiple sites. We can combine prior knowledge with summary statistics to get close to the state-of-the-art, without accessing individual-level data. This can be used for model initialization and also for meta-analysis, prediction, and causality.
✦ Multiple data types. At present we can handle high-dimensional stationary data and longitudinal (time series) data including health records, sensors, telemetry, user-generated data, activity data, web information aggregators, genomics, transcriptomics, metabolomics, proteomics, and glycomics.
Our Team

Dr. Felix Agakov, PhD, CEO. Felix is an
entrepreneur and expert in predictive analytics and machine learning. He has
worked with big pharmas and EU's leading medical universities on biomarker
discovery for chronic diseases, with platform companies on data enrichment
and predictions, and with care providers on risk stratification and
preventive care. He has been the industrial lead in EU-wide consortia and
directed several scientific SMEs.

Prof. Paul McKeigue, PhD, MD, CSO.
Paul is an expert in clinical epidemiology,
statistical genetics, pharmacovigilance, and
Bayesian methods
for genetic epidemiology. He previously held tenured
professorial posts
at the London
School of Hygiene & Tropical Medicine and
University College
Dublin. He is Professor
of statistical genetics in the University of Edinburgh
and Consultant in Public Health to NHS Lothian.

Dr. Peter Orchard, PhD, Senior
Analyst.
Peter is a machine learning expert developing algorithms
and architectures for
network learning and time series modeling. He holds MSci
in physics,
PhD in probabilistic machine learning, and has years of
experience in
high-performance software. His recent focus has been on
deep learning and sparse
methods for telemonitoring and telehealth, and on
high-performance algorithms
for next-generation sequencing.

Dr. Athina Spiliopoulou, PhD, MBA,
Data Scientist.
Athina is an expert in predictive analytics and
business intelligence. She has
worked on predictions using time series and
individual- and summary-level -omics, and she
has been involved in the development of a new
causality analysis tool. She
co-organized summer schools in predictive analytics
and statistical -omics.
She is an expert in communicating the complexities
of data sciences to
non-specialists.

Anna Agakova, MSci, MSci, Data
Engineer.
Anna is an experienced signal processing engineer
leading the developments of signal extraction, data enrichment,
and data integration tools for multiple different
types of data. She has been the lead developer of a new
method for automatic processing of chromatograms
that has resulted in a new service offering. More recently,
she has worked on telemonitoring and on predictions
for non-clinical applications.

Dr. Daniel Urda, PhD, Data
Scientist.
Daniel is an experienced researcher in machine learning
for personalized medicine.
He has worked on a predictive platform for multiple types
of clinical and -omics data,
and co-developed high-dimensional risk stratification
methods. He has been involved in the
development of new risk calculators for rheumatoid
arthritis
and type-2 diabetes, and has worked closely with
specialists in secondary care.

Prof. Chris Haley, PhD,
External SAB.
Chris is an expert in genetic architecture of complex
traits,
methods for identifying major determinants of variation in
complex
traits, animal genetics, and predictions from genome-wide
data.
He is Professor of statistical genetics in the
Roslin Institute and member of Pharmatics's Scientific
Advisory Board.

Dr. Amos Storkey, PhD, External
SAB. Amos
is a recognized expert in probabilistic machine learning,
Bayesian methods,
and applications of data mining. He is developing machine
leaning methods for retinopathy
and renal disease. He is Reader in the School of Informatics
and member of Pharmatics's Scientific Advisory Board.