Explainable Machine Learning Methods for Ocean Worlds Mass Spectrometry Data: Biosignatures and Environmental Characterization

Overview

Development stage

In development or acquisitionSource: a) Pre-deployment – The use case is in a development or acquisition status.

High-impact designation

Not high-impactSource: c) Not high-impact

Impact justification

Not reported

Start date

Not reported

Withheld from public reporting

NoSource: a) No

Mission

Topic area

Science

Operational functions

Clinical and medical Derived by keyword rules; see methodology

Problem the AI is intended to solve

Future astrobiological and geochemical investigations of ocean worlds (OWs) such as Europa and Enceladus will face challenges that can be addressed through science autonomy. While ML methods are potentially powerful tools for the prediction of geochemistry and biosignatures using IRMS data, many of these models are “black boxes” that reduce trust in predictions, and interpretable ML tools are needed to instill trust. In addition, methods are needed to diagnose false predictions for high-stakes predictions like extraterrestrial life.

Expected benefits

We develop and validate an interpretable ML local variable importance tool called Local Nearest-neighbors Projected Distance Regression (local-NPDR) that improves the explainability of ML models for biosignatures and OW chemistry using real IRMS measurements of volatile CO2 from OW analogue brines and simulated data. We hypothesize that false predictions may be identified when the signs and magnitudes of local and global variable importance scores differ. We add local-NPDR false prediction diagnostics to our interpretable ML algorithms that include global-NPDR feature selection and network visualization of globally-important variables. Together these NPDR-based tools add interpretability to ML models with the ability to detect biosignatures and characterize the environment with respect to pH, CO2 concentration and salt content. Such interpretable ML methods will be important for the implementation of science autonomy for future missions to study plumes of OWs and evaluate their habitability and geochemistry.

System outputs

One important example is the use of machine learning (ML) models for the prediction of molecular biosignatures (signals of life) from isotope ratio mass spectrometry (IRMS) data on OW orbiters.

Technology

AI classification

Classical / predictive MLSource: Classical/Predictive Machine Learning: Models trained on data to make predictions or classifications based on identified patterns or relatio…

System name(s)

Not reported

Custom-developed code

Not reported

Public source code

Not reported

Sourcing

How it was built

Not reported

Vendor (as reported)

Not reported

Vendors (standardized)

None on the standard list

A vendor is the supplier named by the agency. It does not identify the underlying model or AI technology.

Data and privacy

Involves PII

Not reported

Privacy Impact Assessment

Not reported

Authorization to Operate

Not reported

Demographic features

Not reported

Training and evaluation data

Not reported

Federal Data Catalog

Not reported

Governance

0 of 8 minimum-practice questions answered. A blank answer means the agency reported nothing; it does not mean the practice is absent.

Potential impacts and how they were identified

Not reported