health · digital · family: it worked in the lab
works on the bench,fails in the field
Multiplexed biosensors that work in the lab fail at point-of-need
Problem statement
Clinical diagnostics increasingly require simultaneous detection of multiple biomarkers — for sepsis, cardiac events, cancer screening, and infectious disease panels — from a single sample at the point of care. Laboratory-based multiplexed assays (e.g., Luminex bead arrays, mass spectrometry) can detect many analytes simultaneously, but these require controlled environments, trained technicians, and sample preparation that is incompatible with field or bedside use. Despite continued research into miniaturized multiplexed biosensors, reliable detection of many analytes at once in unprocessed clinical samples (whole blood, saliva, urine) outside a laboratory setting remains an open research need.
Why this matters
Sepsis is one example: earlier detection of its multi-biomarker signature (procalcitonin, lactate, IL-6, C-reactive protein) could help guide treatment sooner. Rapid multi-pathogen detection panels for respiratory illness (COVID, influenza, RSV, bacterial pneumonia) would transform emergency department triage. In low-resource settings, a single multiplexed rapid test could replace multiple individual lateral flow assays, reducing cost and sample volume requirements. The inability to perform multiplexed detection at point-of-need means that diagnostic information arrives too late to guide initial treatment decisions.
What’s been tried and why it hasn’t worked
Microfluidic lab-on-a-chip devices can integrate multiple detection zones but suffer from cross-reactivity between closely spaced assays, bubble formation that disrupts flow, and clogging from complex biological matrices like whole blood. Electrochemical sensor arrays using functionalized electrodes achieve multiplexing in buffer solutions but experience fouling and signal drift on exposure to real clinical samples. Paper-based lateral flow assays can be multiplexed to a few analytes but lack quantitative accuracy and dynamic range. Surface-enhanced Raman scattering (SERS) platforms offer molecular fingerprinting but require expensive laser systems and are sensitive to temperature fluctuations. The fundamental challenge is that strategies to increase analytical sensitivity (more surface area, more capture molecules, longer incubation) conflict with strategies needed for rapid, simple, field-robust operation.
What would unlock progress
A breakthrough in anti-fouling surface chemistry that maintains biorecognition element activity in complex samples for the duration of the assay would address the most persistent failure mode. Alternatively, sample preparation approaches that can separate plasma from whole blood in seconds without external equipment (passive microfluidics, acoustic separation) would make existing multiplexed detection technologies viable. Machine learning approaches to deconvolve overlapping signals from cross-reactive sensor arrays could circumvent the physical separation requirement entirely.
Entry points for student teams
A student team could build a benchtop prototype that combines a simple passive plasma separation membrane with a 4-plex electrochemical detection array, testing it against spiked whole blood samples and comparing performance to laboratory reference methods. Alternatively, a team could develop and validate an ML-based signal deconvolution algorithm for a commercial electrochemical sensor array exposed to mixed analyte solutions. Relevant disciplines include biomedical engineering, analytical chemistry, and machine learning.
Genes
Structural cousins: stuck the same way, in other fields
Sources
NSF CBET Biosensing Program and Nano-Biosensing Program, Division of Chemical, Bioengineering, Environmental and Transport Systems; accessed 2026-02-15 go to source ↗
Working notes
Notes made while writing and checking this brief, shown as written. They use working shorthand.
The NSF Biosensing program explicitly encourages proposals addressing "critical sensor needs in biomedical research, public health, food safety, agriculture, forensics, environmental protection, and homeland security." The Nano-Biosensing program highlights the need for "multi-purpose sensor platforms that exceed the performance of current state-of-the-art measurement methods." Related problem in collection: health-longterm-implantable-glucose-sensor.md shares the biofouling barrier but in a single-analyte, implantable context. The multiplexed field-translation problem is fundamentally different because it involves managing cross-reactivity and complex sample matrices in a rapid-use format.
Reconciliation 2026-09-29: A site review read this write-up against both NSF pages it names. Both were fetched in full on 2026-09-29: the Biosensing page (PD 20-7909) from https://www.nsf.gov/funding/opportunities/biosensing, and the Nano-Biosensing page (PD 15-7909) from https://www.nsf.gov/funding/opportunities/nano-biosensing. Both pages are short program descriptions and NSF now marks both as archived. Neither page contains any figure about sepsis, about how many analytes a sensor can detect, or about how fast sensors fail. Figures that are not on either page were taken out or softened rather than re-sourced. (1) The text said no platform reliably detects "more than 3-4 analytes" in unprocessed samples outside a laboratory, after "two decades of research." Neither page says this. The sentence now says that reliable detection of many analytes at once in unprocessed samples outside a laboratory remains an open research need. (2) The text said laboratory assays detect "dozens of analytes"; it now says "many." (3) The text said "Sepsis kills approximately 350,000 Americans annually" and that detection "within the first hour of presentation can reduce mortality by 30-40%." Neither page mentions sepsis. The sentence now gives sepsis as an example where earlier detection could help guide treatment sooner, with no figures. (4) The text said electrode arrays foul and drift "within minutes" and that paper-based tests can be multiplexed "to 2-3 analytes." Neither figure is on either page. The sentences now say "on exposure" and "a few analytes." (5) The Source Notes quoted the Nano-Biosensing page as "multi-purpose sensor platforms that exceed the performance of current state-of-the-art devices." The page says "measurement methods," not "devices"; the quotation now matches the page. Confirmed against the pages and left as written: the Source line (both programs sit in the Division of Chemical, Bioengineering, Environmental and Transport Systems); the quoted list of sensor needs from the Biosensing page, which is word for word; and the general direction of the problem, since the Biosensing page names "adaptive and/or multiplex sensing technologies" as an area of critical need and lists combining biosensors with artificial intelligence, and the Nano-Biosensing page lists measurement in practical samples without sample preparation and the stability of recognition molecules at sensor surfaces. NOT supported by either page and left as written because they carry no figures: the descriptions of how each sensor type fails (cross-reactivity, bubbles and clogging in microfluidic chips, fouling of electrodes, the cost and temperature sensitivity of Raman systems), the named laboratory methods, the four sepsis biomarkers, and the claims about emergency triage and low-resource settings. These are general background and need a technical source of their own. One point for readers who follow the link: the Biosensing page says the program does not encourage proposals on "microfluidics for sample separation and filtration," so the sample-separation direction in "What Would Unlock Progress" is one that program did not encourage. The same page lists combining biosensors with artificial intelligence as an area of interest, and sends proposals on sensing systems that use machine learning to a different NSF program. Genes untouched.