Industry Expertise

Life Sciences & Healthcare

Enhance Discoveries | Accelerate Potential

For life sciences and health care companies, we operate at the intersection of biology and data science. We are a team of data science experts partnering with early-stage biotech, biopharma, and medical device companies as consultants or as a risk partner to integrate analytics and machine learning capabilities to improve care, enhance discoveries and enable data-driven decisions. Our multidisciplinary scientists are experts in computational biology, genomics, bioinformatics, pharmacology, neuroscience, biophysics, and data science.

Build an AI platform to find targets of genetic engineering to improve crop yield

Analyze plant genomic data and recommend genetic engineering targets to increase yield and resilience of important, staple food crops using a scalable AI platform.

Read more

Build an AI platform to find targets of genetic engineering to improve crop yield

Analyze plant genomic data and recommend genetic engineering targets to increase yield and resilience of important, staple food crops using a scalable AI platform.

The Challenge

Our agricultural biotech client needed a platform to identify gene targets for desired traits from complex genomic and environmental relationships, based on genome wide association studies (GWAS), and in context of current scientific knowledge.

Outcome

We used machine learning to identify the association of genomic variants to plant traits. We trained NLP models on a large set of published scientific literature to put variant recommendations in the context of global biomedical knowledge, providing scientists with a better understanding of past studies and the competitive landscape. Early results on crop yields have proven the validity of causal gene recommendations.

Read more
Build an AI platform to find targets of genetic engineering to improve crop yield
The Challenge

Our agricultural biotech client needed a platform to identify gene targets for desired traits from complex genomic and environmental relationships, based on genome wide association studies (GWAS), and in context of current scientific knowledge.

Our Solution

We used machine learning to identify the association of genomic variants to plant traits. We trained NLP models on a large set of published scientific literature to put variant recommendations in the context of global biomedical knowledge, providing scientists with a better understanding of past studies and the competitive landscape. Early results on crop yields have proven the validity of causal gene recommendations.

Create a digital diagnostics platform for behavioral health

A better way to diagnose patients with depression and anxiety for a company bringing new therapeutics to market for neurobehavioral disorders

Read more

Create a digital diagnostics platform for behavioral health

A better way to diagnose patients with depression and anxiety for a company bringing new therapeutics to market for neurobehavioral disorders

The Challenge

Our client is bringing new therapeutics to market for neurobehavioral disorders and needed a better way to diagnose patients with depression and anxiety based on video and audio content. A platform was needed to enable a digital diagnosis to support patient stratification for clinical trials and assess the efficacy of treatment.

Outcome

We partnered with their data science and software engineering teams to design and build a platform to use machine learning to improve the diagnosis of patients with depression and anxiety. Time series audio and video data, captured via a mobile app, were feature engineered to train a model for predicting clinical outcomes. Improved diagnosis will enhance the stratification of patients for trial enrollment and accelerated assessment of therapeutic response.

Read more
Create a digital diagnostics platform for behavioral health
The Challenge

Our client is bringing new therapeutics to market for neurobehavioral disorders and needed a better way to diagnose patients with depression and anxiety based on video and audio content. A platform was needed to enable a digital diagnosis to support patient stratification for clinical trials and assess the efficacy of treatment.

Our Solution

We partnered with their data science and software engineering teams to design and build a platform to use machine learning to improve the diagnosis of patients with depression and anxiety. Time series audio and video data, captured via a mobile app, were feature engineered to train a model for predicting clinical outcomes. Improved diagnosis will enhance the stratification of patients for trial enrollment and accelerated assessment of therapeutic response.

Identify a drug’s mechanism of action by integrative analysis of proteomics

Analyze proteomics data to determine the MOA of a novel drug candidate that showed promise in in-vitro and in-vivo studies

Read more

Identify a drug’s mechanism of action by integrative analysis of proteomics

Analyze proteomics data to determine the MOA of a novel drug candidate that showed promise in in-vitro and in-vivo studies

The Challenge

Our client, a pulmonary therapeutics company, has a novel drug candidate showing promise in both in vitro and in vivo studies. However, the drug’s mechanism of action was not well understood, leading to uncertainty and risk for future clinical trials. Our client conducted a functional proteomics assay (RPPA) on diseased, healthy, and tissues treated with their drug candidate. They needed a data science-driven analysis of the RPPA data to identify the genes, proteins, and pathways involved in both the diseased and treated state.

Outcome

Our team used data science techniques to find clusters of proteins up or down-regulated in the disease state and in the treated state relative to healthy states. We mapped patterns of protein levels to pathways to infer mechanism of action of treatment. We confirmed hypotheses to guide the next phase of development for their drug. The clients used and sourced RPPA, an antibody-based assay, from an academic core facility with highly validated antibodies. This work holds additional promise for developing a clinical assay to determine therapeutic response during trials.

Read more
Identify a drug’s mechanism of action by integrative analysis of proteomics
The Challenge

Our client, a pulmonary therapeutics company, has a novel drug candidate showing promise in both in vitro and in vivo studies. However, the drug’s mechanism of action was not well understood, leading to uncertainty and risk for future clinical trials. Our client conducted a functional proteomics assay (RPPA) on diseased, healthy, and tissues treated with their drug candidate. They needed a data science-driven analysis of the RPPA data to identify the genes, proteins, and pathways involved in both the diseased and treated state.

Our Solution

Our team used data science techniques to find clusters of proteins up or down-regulated in the disease state and in the treated state relative to healthy states. We mapped patterns of protein levels to pathways to infer mechanism of action of treatment. We confirmed hypotheses to guide the next phase of development for their drug. The clients used and sourced RPPA, an antibody-based assay, from an academic core facility with highly validated antibodies. This work holds additional promise for developing a clinical assay to determine therapeutic response during trials.

Automate the discovery of advanced nutritional products

Determine the MOA of a new drug lead to support the development of AI drug discovery techniques to identify similar compounds

Read more

Automate the discovery of advanced nutritional products

Determine the MOA of a new drug lead to support the development of AI drug discovery techniques to identify similar compounds

The Challenge

Our client is working to identify nutritional based supplements to improve health. Our client has a drug candidate showing promise in improving cardiovascular health and endurance based on in vivo animal studies. However, the specific mechanism of action was not well understood.

Outcome

We designed a study and protein expression analysis was done using reverse-phase protein lysate microarray (RPPA) from an academic core facility. Our team was able to identify patterns of protein activity and infer mechanism of action. This work makes possible the development of a screening platform to identify future compounds with a similar profile.

Read more
Automate the discovery of advanced nutritional products
The Challenge

Our client is working to identify nutritional based supplements to improve health. Our client has a drug candidate showing promise in improving cardiovascular health and endurance based on in vivo animal studies. However, the specific mechanism of action was not well understood.

Our Solution

We designed a study and protein expression analysis was done using reverse-phase protein lysate microarray (RPPA) from an academic core facility. Our team was able to identify patterns of protein activity and infer mechanism of action. This work makes possible the development of a screening platform to identify future compounds with a similar profile.

A better way to prioritize drug targets

Evaluate large libraries of therapeutic candidates against published literature, patents and clinical trials to prioritize the best targets for drug development

Read more

A better way to prioritize drug targets

Evaluate large libraries of therapeutic candidates against published literature, patents and clinical trials to prioritize the best targets for drug development

The Challenge

A cancer therapeutics company has a high volume platform to identify novel drug targets. Once identified, their scientists need to know if a target is, novel, if it has been previously studied, has been tested in a prior clinical trial, and if it is the subject of any patent activity. The challenge was to create a data-driven process to prioritize which therapeutic candidates should move forward in the drug development pipeline. A platform is needed to analyze large libraries of therapeutic candidates against the body of scientific literature and public data to provide context and insights.

Outcome

Our team trained NLP models to extract gene to disease relationships from ~2 TB of scientific literature, clinical trials, and patents to create a target evaluation platform, enabling their scientists to prioritize the highest quality targets for further research. As a result, our clients were able to sign collaboration agreements with new pharma partners based on the quality of the drug targets as highlighted in the data platform combined with pre-clinical data.

Read more
A better way to prioritize drug targets
The Challenge

A cancer therapeutics company has a high volume platform to identify novel drug targets. Once identified, their scientists need to know if a target is, novel, if it has been previously studied, has been tested in a prior clinical trial, and if it is the subject of any patent activity. The challenge was to create a data-driven process to prioritize which therapeutic candidates should move forward in the drug development pipeline. A platform is needed to analyze large libraries of therapeutic candidates against the body of scientific literature and public data to provide context and insights.

Our Solution

Our team trained NLP models to extract gene to disease relationships from ~2 TB of scientific literature, clinical trials, and patents to create a target evaluation platform, enabling their scientists to prioritize the highest quality targets for further research. As a result, our clients were able to sign collaboration agreements with new pharma partners based on the quality of the drug targets as highlighted in the data platform combined with pre-clinical data.

Guide medical diagnosis with biometric eye tracking

Design and build a realtime gaze tracking application to assess cognitive performance in dementia, concussion, and intoxication patient cohorts

Read more

Guide medical diagnosis with biometric eye tracking

Design and build a realtime gaze tracking application to assess cognitive performance in dementia, concussion, and intoxication patient cohorts

The Challenge

Our client needed to assess gaze tracking as part of their cognitive assessment platform. Specifically, they needed a mobile application to assess cognitive functioning in realtime for healthy vs. not-healthy behavior in dementia, concussion, and intoxication patient cohorts. 

Outcome

We designed a tablet-based application to record a person’s ability to track a visual stimulus (nystagmogram) for healthy, non-concussed individuals. We conducted 2D and 3D feature engineering of the videos to extract facial landmarks, including corrections for head position, orientation, and angle. We trained a convolutional neural net to learn the focus of the gaze, and score a new individual’s ability to track a visual stimulus. Performance-tuned models delivered responses in realtime within six degrees of accuracy on a mobile platform.

Read more
Guide medical diagnosis with biometric eye tracking
The Challenge

Our client needed to assess gaze tracking as part of their cognitive assessment platform. Specifically, they needed a mobile application to assess cognitive functioning in realtime for healthy vs. not-healthy behavior in dementia, concussion, and intoxication patient cohorts. 

Our Solution

We designed a tablet-based application to record a person’s ability to track a visual stimulus (nystagmogram) for healthy, non-concussed individuals. We conducted 2D and 3D feature engineering of the videos to extract facial landmarks, including corrections for head position, orientation, and angle. We trained a convolutional neural net to learn the focus of the gaze, and score a new individual’s ability to track a visual stimulus. Performance-tuned models delivered responses in realtime within six degrees of accuracy on a mobile platform.

Strategic consulting to build a scalable data engineering team

Create a strategy to build a data engineering team to support an AI-driven agriculture biotech's rapid growth

Read more

Strategic consulting to build a scalable data engineering team

Create a strategy to build a data engineering team to support an AI-driven agriculture biotech's rapid growth

The Challenge

Our client is at the forefront of innovation in agricultural biotech, working to increase the yield of food crops. They are scaling rapidly and needed interim engineering leadership to scale their team to support the company’s evolution. The company’s software platforms were facing an order of magnitude increase in available data to drive R&D. They needed strategic advice to create the right team using the right tools and following a proven strategy to automate data engineering pipelines and unlock the value of new data.

Outcome

We worked with our client to set engineering standards, create a framework for engineering to fit within the business and establish longer-term technology strategies that can evolve with the organization. Our client is now recruiting to build the data engineering team to scale for their future.

Read more
Strategic consulting to build a scalable data engineering team
The Challenge

Our client is at the forefront of innovation in agricultural biotech, working to increase the yield of food crops. They are scaling rapidly and needed interim engineering leadership to scale their team to support the company’s evolution. The company’s software platforms were facing an order of magnitude increase in available data to drive R&D. They needed strategic advice to create the right team using the right tools and following a proven strategy to automate data engineering pipelines and unlock the value of new data.

Our Solution

We worked with our client to set engineering standards, create a framework for engineering to fit within the business and establish longer-term technology strategies that can evolve with the organization. Our client is now recruiting to build the data engineering team to scale for their future.

Accelerate oncology drug discovery by analyzing public proteomic data

Integrate public proteomic data to enhance and validate antibody screening results

Read more

Accelerate oncology drug discovery by analyzing public proteomic data

Integrate public proteomic data to enhance and validate antibody screening results

The Challenge

Our client had screening assay results for a large antibody panel and protein expression profiles collected from several human tissues. The results were promising, but additional validation was needed.

Outcome

We obtained publicly available proteomic data presented as IHC stained digital pathology slides of cancer and healthy tissues. Using computer vision, we were able to identify a set of proteins highly expressed in cancer tissues vs. healthy tissues, and translate these into biomarkers for each type of cancer. This analysis provided our client with the additional insight and validation required to continue their drug discovery efforts.

Read more
Accelerate oncology drug discovery by analyzing public proteomic data
The Challenge

Our client had screening assay results for a large antibody panel and protein expression profiles collected from several human tissues. The results were promising, but additional validation was needed.

Our Solution

We obtained publicly available proteomic data presented as IHC stained digital pathology slides of cancer and healthy tissues. Using computer vision, we were able to identify a set of proteins highly expressed in cancer tissues vs. healthy tissues, and translate these into biomarkers for each type of cancer. This analysis provided our client with the additional insight and validation required to continue their drug discovery efforts.

Predictive modeling to identify mechanical biomarkers to accelerate oncology diagnosis

Machine learning to maximize signal from noise to differentiate normal vs malignant tissue

Read more

Predictive modeling to identify mechanical biomarkers to accelerate oncology diagnosis

Machine learning to maximize signal from noise to differentiate normal vs malignant tissue

The Challenge

Our client has a nano-mechanical sensor in clinical trials that is able to differentiate cell types within patient biopsies. This innovation promises to accelerate the diagnosis of cancer and offer personalized treatment recommendations. Our client needed data science-driven analysis of the machine signal to identify mechanical biomarkers to differentiate normal and malignant tissue.

Outcome

We feature engineered the machine signal and applied clustering, regression and machine learning analysis. We built a data engineering pipeline to handle large scale measurement data. We trained a predictive model to differentiate between normal and malignant cells, providing insight into the most important mechanical biomarkers. We are now assisting our client to build their data science team.

Read more
Predictive modeling to identify mechanical biomarkers to accelerate oncology diagnosis
The Challenge

Our client has a nano-mechanical sensor in clinical trials that is able to differentiate cell types within patient biopsies. This innovation promises to accelerate the diagnosis of cancer and offer personalized treatment recommendations. Our client needed data science-driven analysis of the machine signal to identify mechanical biomarkers to differentiate normal and malignant tissue.

Our Solution

We feature engineered the machine signal and applied clustering, regression and machine learning analysis. We built a data engineering pipeline to handle large scale measurement data. We trained a predictive model to differentiate between normal and malignant cells, providing insight into the most important mechanical biomarkers. We are now assisting our client to build their data science team.

Automated quality assurance for medical image (MRI) data sets

Automatically evaluate which MRI scans can be used for subsequent analysis and modeling

Read more

Automated quality assurance for medical image (MRI) data sets

Automatically evaluate which MRI scans can be used for subsequent analysis and modeling

The Challenge

Our client needs to train predictive models based on large sets of MRIs, sourced from many clinical locations, and public data sets. Manual quality assurance to determine which images met the criteria for the study were not scalable. Our client needed a fully automated solution to determine which images met the quality threshold.

Outcome

We created an automated quality control classifier to determine which MRI scans were allowed to proceed to the next level of analysis. Cross-study performance analysis confirmed the generalizability of the model. Further, our data pipeline extracts features from the scan process logs using a random-forest classification algorithm for the predictive modeling phase.

Read more
Automated quality assurance for medical image (MRI) data sets
The Challenge

Our client needs to train predictive models based on large sets of MRIs, sourced from many clinical locations, and public data sets. Manual quality assurance to determine which images met the criteria for the study were not scalable. Our client needed a fully automated solution to determine which images met the quality threshold.

Our Solution

We created an automated quality control classifier to determine which MRI scans were allowed to proceed to the next level of analysis. Cross-study performance analysis confirmed the generalizability of the model. Further, our data pipeline extracts features from the scan process logs using a random-forest classification algorithm for the predictive modeling phase.