418 9.4  Personalizing Healthcare

For many microcantilever systems, changes of ~1 in a 1000 in v0 can be measured across

the bandwidth range, equating to ~1 Hz that corresponds to a change in mass of 10⁻¹⁵ kg

(or 1 picogram, pg). This may seem a small amount, but even for a large protein of ~100

kDa molecular weight, this is equivalent to the binding of ~10⁶ molecules. However, a real

advantage with this method involves using multiple cantilevers with different antibody

coatings inside the chip device to enable a signature for the presence of a range of different

biomolecules present in a sample to be built up. Improvements in high-​throughput and sig­

nature detection can be made using a similar array strategy of multiple detection zones using

other biophysical detection techniques beyond microcantilevers.

Mechanical signals are also emerging as valuable metrics for cell types in biosensing,

for example, using AFM to probe the stiffness of cell membranes, and also, optical force

tools such as the optical stretcher to measure cell elasticity (see Chapter 6). Cell mechanics

change in disease states, though for cancer the dependence is complex since different types

of cancers can result in either increasing or decreasing the cell stiffness and also may have

different stiffness values at different stages in tumor formation.

The faithful interpretation of relatively small biomolecule signals from portable lab-​

on-​a-​chip devices presents problems in terms of poor stability and low signal-​to-​noise

ratios. Improvements in interpretation can be made using smart computational inference

tools, for example, Bayesian inference (see Chapter 8). Although signals from individual

biomarkers can be noisy, integrating the combination of multiple signals from different

biomarkers through Bayesian inference can lead to greatly improved fidelity of detection.

Much of this computation can be done decoupled from the hardware of the device itself,

for example, to utilize smartphone technology. This enables a reduction in both the size

and cost of the device and makes the prospect of such devices emerging into clinics in the

near future more of a reality.

A Holy Grail in the biosensing field is the ability to efficiently and cheaply sequence single

molecules of DNA. For example, the so-​called $1000 genome refers to a challenge set for

sequencing technologies called next-​generation sequencers, which combine several biophys­

ical techniques to deliver a whole genome sequence for a cost of $1000 or less. The U.S. biotech

company Illumina has such a machine that purports to do so, essentially by first fragmenting

the DNA, binding the fragments to a glass slide surface, amplifying the fragments using PCR

(see Chapter 7), and detecting different fluorescently labeled nucleotide bases tagged with

four different dyes in these surface clusters. Bioinformatics algorithms are used to compu­

tationally stitch the sequenced fragments together to predict the full DNA sequence. This is

similar to the original Sanger shotgun approach for DNA sequencing, which generated DNA

fragments but then quantified sequence differences by running the fragments on gel elec­

trophoresis. However, the combination of amplification, clustering, and fluorescence-​based

detection results in far greater high throughput (e.g., one human genome sequenced in a

few days). However, the error rate is ca. 1 in 1000 nucleotides, which may seem small but in

fact results in 4 million incorrectly predicted bases in a typical genome, so there is scope for

improvement.

A promising new type of sequencing technology uses ion conductance measurements

through engineered nanopores, either solid state of manufactured from protein adapters

such as α-​hemolysin (see Chapter 6). For example, in 2012, the UK biotech company Oxford

Nanopore released a disposable sequencing device using such technology that could interface

with a PC via a USB port and which cost less than $1000. The device had a stated accuracy

of ~4%, which makes it not sufficiently precise for some applications, but even so with a cap­

ability to sequence DNA segments up to ~1 million base pairs in length, this represents a

significant step forward.

A promising area of smart in vivo diagnostics involves the use of synthetic biological

circuits inside living cells. For example, such methods can in principle turn a cell into a bio­

logical computer that can detect changes to its environment, record such changes in memory

composed of DNA, and then stimulate an appropriate cellular response, for example, send a

signal to stop a cell from producing a particular hormone and/​or produce more of another

hormone, or to stimulate apoptosis (i.e., programmed cell death) if a cancer is detected.