Our revolutionary method in oil and gas exploration uses high-end computing and machine learning to analyse hundreds of thousands of bacterial species


Soil / Seabed samples

Samples taken by surveyors, < 1m below seabed with simple drop coring (offshore), or our own field crew (onshore sampling, 1 foot below the surface).
  • Our patented technology – developed to analyze surface soil or seabed samples (few mm3’s, from approx. 1 ft. depth) – recognizes otherwise undetectable hydrocarbon microseepage from prospective areas.

Field crew will travel to the location to sample the area of interest, either in a grid pattern (covering a lease or area of interest) or visiting different target locations.

The samples are taken with specialized equipment and are stored in hermetic conditions to avoid contamination.

DNA analysis

Bacterial DNA is extracted, producing tagged 16S rDNA, translated to bacterial species, thus creating a unique ‘DNA fingerprint’ of the microbial ecosystem of the soil sample.
  • After the sample has been taken, first bacterial DNA is extracted from this sample, producing tagged 16S rDNA data that is translated into bacterial species.
  • The result of this analysis is hundreds of thousands (partly field type specific) biomarkers. The ‘DNA fingerprint’ of the soil sample.
  • Biodentify creates a prospectivity ‘indicator’ on the sampled new locations using the DNA fingerprints, which is a value between -1 and 1. These values are plotted on a map which shows the prospect areas very distinctively and along on cross-section.

Schematic representation of the bacterial diversity and relative abundance within samples. Each colored bar represents a bacterial genus or family.

One soil sample could contain up to 35 million DNA signatures and 350,000 species. The extraction of microsomal chromosomal DNA is performed by mechanical disruption of the bacteria using small beads and vigorous shaking.

Machine learning

Our existing database with known samples (taken over oil/gas reservoirs, and dry wells), a correlation model is built and predictions for the new samples can be made.
  • Our database – with over 5400 samples from both onshore and offshore locations, detailed ‘DNA fingerprints’, with related production data on biomarkers – is subsequently used as modeling input to our proprietary localized triple loop © computational model.
  • Our trained model is capable of predicting the presence of a highly productive zone within a heterogeneous shale play, with an accuracy of >70% prior to drilling. For conventional plays, both on- and offshore, it is able to predict whether a prospect or well target location shows signs of hydrocarbons in the subsurface.
  • Our model uses the difference in DNA fingerprints that are connected to highly productive areas and those correlated to unproductive (or dry) areas or locations.

Our machine learning model finds the ‘signal in the noise’. Only when enough biomarkers (50-200) are combined, a sufficiently accurate estimate can be produced on the relative prospectivity. The algorithms using machine learning to find a small difference in the DNA fingerprints related to microseepage and they neglect the influences of the climate, local spills or other.

Prospect map

Model is used to de-risk prospects, creating a predictive map where prospects or well targets are ranked in order of prospectivity.
  • Six plays in the US are extensively sampled, with different characteristics on productivity, the age of the producing interval, type of play (oil vs gas), geology and climate and soil type: Bone Spring (Permian), Bakken, Antrim, Avalon, Lewis, Haynesville, and Marcellus. In addition, studies have been carried out both on- and offshore the Netherlands as well as offshore Norway.

Final deliverable: prospectivity map, highlighting highly productive zone (in red) versus low producing zone (blue) – prior to drilling, based only on shallow soil samples.

Predictions for the new samples can be made with an accuracy between 72-86% (based on US Onshore benchmark study).

Our Publications