RNA melting baseline

An RNA melting baseline is the predicted absorbance of a specified folded or unfolded RNA population as a function of temperature, before mixing those populations in an RNA optical melting model. It differs from a solvent blank: a folded RNA sample still absorbs light, and that response can change with temperature even without the modeled transition.

For a two-state RNA melting model with unfolded fraction $f_U$, writeThe folded and unfolded RNA melting baselines might be modeled as $A_F=a_F+b_{F}T$ and $A_U=a_U+b_{U}T$. Their continuation into the transition region is RNA baseline extrapolation. Solving the optical mixture givesFor original stipulated readings $A=0.55$, $A_F=0.40$ and $A_U=0.70$, the inferred fraction is 0.50. Changing only the folded RNA melting baseline to 0.45 gives 0.40; changing only the unfolded RNA melting baseline to 0.80 gives 0.375. The same measured absorbance can therefore imply different populations under different baseline models.

Sensitivity is explicit. With $D=A_U-A_F$,A small baseline separation amplifies measurement uncertainty. Correlated fitted baselines require covariance terms in uncertainty propagation, rather than independent error bars by assumption.^[ref]

Optical-melting studies infer thermodynamics by fitting measured curves, rather than directly counting unfolded molecules.^[ref] Inspect the pre-transition and post-transition regions and test the chosen RNA baseline extrapolation. An inferred fraction outside $[0,1]$ signals noise or incompatibility with the assumed optical model; it is not a negative molecular population.

RNA baseline extrapolation continues a fitted folded-state or unfolded-state RNA melting baseline from its observed temperature range into the transition region. It is necessary when estimating populations from RNA optical melting, because both hypothetical pure-state signals are needed at temperatures where the measured sample contains a mixture.

For a linear folded baseline, choose a reference $T_0$ and writeThe reference merely changes parameterization. An original fit with $T_0=300\mathrm{K}$, $a=0.400$ and $b=0.001{\mathrm{K}}^{-1}$ predicts 0.420 at $320\mathrm{K}$ and 0.450 at $350\mathrm{K}$. The first prediction might interpolate within measured data; the second extrapolates if all folded-state observations were below $320\mathrm{K}$.

The inferred population depends on that continuation. At $350\mathrm{K}$, suppose measured absorbance is 0.600 and the unfolded RNA melting baseline is 0.700. The linear folded-baseline prediction givesIf the actual pure folded-state response there were 0.490, the same optical-mixture equation would instead give $f_U\approx 0.524$. A good fit in a limited low-temperature region does not rule out curvature beyond it.

RNA baseline prediction covariance is important even within the chosen linear model. With $x=T-T_0$,For stipulated independent fit parameters $u_a=0.002$ and $u_b=0.0002{\mathrm{K}}^{-1}$, the prediction's standard uncertainty rises from 0.002 at $x=0$ to approximately 0.0102 at $x=50\mathrm{K}$. This is uncertainty propagation for estimated parameters; it does not include unmodeled curvature.^[ref]

Optical-melting experiments fit signal curves to infer thermodynamics.^[ref] Report which regions constrain each RNA melting baseline, inspect residuals and compare plausible baseline forms. Confidence in the fitted parameters and confidence in the functional form are different parts of the inference.

RNA baseline prediction covariance quantifies how errors in fitted RNA melting baselines link their predicted values at different temperatures. Reusing the same fit parameters creates shared uncertainty; predictions are not independent merely because they are evaluated at different temperatures.

For a linear RNA baseline extrapolation $A_F(T)=a+b(T-T_0)$, put $x_i=T_i-T_0$. ThenThis relation follows from the covariance of two linear combinations. The diagonal $C_{ii}$ gives a prediction variance; the off-diagonal entries determine how predictions move together.^[ref]

An original example assumes independent parameters with $u_a=0.002$ and $u_b=0.0002{\mathrm{K}}^{-1}$. At temperatures $20\mathrm{K}$ below and above $T_0$, both prediction variances are $20\times 10^{-6}$, while their covariance is $-12\times 10^{-6}$. The correlation coefficient is therefore $-0.6$. A slope error moves the predictions in opposite directions around the reference point, outweighing the common intercept error.

The difference between those two predictions cancels the intercept exactly:Its standard uncertainty is 0.008. Ignoring covariance would give $\sqrt{40\times 10^{-6}}\approx 0.00632$, which is too small for this model. Shared fitted inputs can produce negative as well as positive prediction correlations.

In vector form, a prediction covariance matrix is $C=X{\Sigma }_{\theta }{X}^{\mathsf{T}}$, with rows of $X$ equal to $(1,x_i)$ and parameter covariance ${\Sigma }_{\theta }$. A two-parameter fit supplies at most two independent shared-error directions, even if evaluated at a hundred temperatures. Additional independent readout errors contribute their own covariance.

When these predictions are used to infer populations in RNA optical melting, propagate their joint uncertainty rather than only separate error bars. This accounting does not test whether a linear baseline is chemically adequate; curvature or an incorrect pure-state model needs separate evidence.