DATA DRIVEN DECISION MAKING RYAN RAMANUJAM … · Dr. Pierre Dönnes - PhD in Bioinformatics - Established business development record in big pharma Prof. Wojciech Chachólski - Professor,

DATA DRIVEN DECISION MAKING

RYAN RAMANUJAM

[email protected]

Current methods: Machine learning, AI, classical statistics

Often insufficient to analyze data due to noise, inaccurate measurements

DATAVERGE PIPELINE

Data Sources ETL GEODE REPORTS

Li et al. Sci Transl Med (2015)

TOPOLOGICAL DATA ANALYSIS

INDUSTRIES

Life Sciences – how can treatments be optimized for cancer

patients?

Finance – which factors influence predictive currency

exchange rates?

Manufacturing/production – which car parts fail early and

under which conditions?

Marketing/sales – which products and strategies will boost

sales?

GEOMETRIC DATA EXPLORER (GEODE)

Pre-processing

Data partitioning

Comparison of expression data from gastric cancer

- Hippo, et. Al. “Global Gene Expression Analysis of Gastric Cancer by

Oligonucleotide Microarrays”. Cancer Research (2002)

n=30, 22 gastric cancer tissues and 8 noncancerous tissue

6936 genes measured

Goals:

1) Correctly classify samples (cross-validation)

2) Determine which genes have highest correlation with disease status

3) Generate subgroups for doctors to determine best patient treatment

GASTRIC CANCER EXAMPLE

GEO GRAPHING

Prediction accuracy: GEODE 94.8%

SVM classifier 72.4%

GEO SIGNATURES

Subgroup 1• Gene X05276_at upregulated

• Gene X12791_at downregulared

Subgroup 2• Gene K01396_at upregulated

• Gene D14663_at upregulated

• Gene U90426_at upregulated

Subgroup 3• Gene M33308_at downregulated

• Gene X14008_rna1_f_at downregulated

GASTRIC CANCER EXAMPLE

DATAVERGE USE CASE

Stage 1 – Customer interaction

DATAVERGE USE CASE

Stage 2 – GEODE Analysis

DATAVERGE USE CASE

Stage 2 – GEODE Analysis

DATAVERGE USE CASE

Stage 2 – GEODE Analysis

Model building

DATAVERGE USE CASE

Stage 3 – Reporting

DATAVERGE USE CASE

Stage 3 – Reporting

Dr. Ryan Ramanujam- Associate Prof. Karolinska Institutet, Researcher KTH

- MBA IE Business School, 23 biomedical publications

Dr. Pierre Dönnes- PhD in Bioinformatics

- Established business development record in big pharma

Prof. Wojciech Chachólski- Professor, KTH Dept. of Mathematics

- World leading expert in algebraic topology

DATAVERGE TEAM

We provide new ways to analyze data, giving better

accuracy and interpretation of results

The underlying algorithms are based on research done

at KTH and Karolinska Institutet

Early support provided by the KTH Innovation office

and a 2017 VINNOVA Innovation Startup Grant

We are looking for partners who want to fully harness the

power of their data

DATAVERGE SUMMARY

WE HAVE THE METHODS

NOW WE NEED YOUR DATA!

[email protected]