When running molecular dynamics (MD) simulations, we are usually interested in measuring an ensemble average of some metric (e.g., RMSD, RMSF, radius of gyration, …) and use this to draw conclusions about the investigated system. While calculating the average value of a metric is straightforward (we can simply measure the metric in each frame and average it) calculating a statistical uncertainty is a little more tricky and often forgotten. The main challange when trying to calculate an uncertainty of MD oveservables is that individual frames of the simulation are not samped independently but they are time correlated (i.e., frame N depends on frame N-1). In this blog post, I will breifly introduce block averaging, a statistical technique to estimate uncertainty in correlated data.
Continue readingCategory Archives: Statistics
Combining Multiple Comparisons Similarity plots for statistical tests
Following on from my previous blopig post, Garrett gave the very helpful suggestion of combining Multiple Comparisons Similarity (MCSim) plots to reduce information redundancy. For example, this an MCSim plot from my previous blog post:
This plot shows effect sizes from a statistical test (specifically Tukey HSD) between mean absolute error (MAE) scores for different molecular featurization methods on a benchmark dataset. Red shows that the method on the y-axis has a greater average MAE score than the method on the x-axis; blue shows the inverse. There is redundancy in this plot, as the same information is displayed in both the upper and lower triangles. Instead, we could plot both the effect size and the p-values from a test on the same MCSim.
ggPlotting tips with OPIG data
Ever wondered whether opiglets keep their ketchup in the fridge or cupboard? Perhaps you’ve wanted to know how to create nice figure to display lots of information simulataniously. Publication quality figures are easy in R with the ggplot package. We may also learn some good visualisation.
Continue readingThe War of the Roses: Tea Edition
Picture the following: the year is 1923, and it’s a sunny afternoon at a posh garden party in Cambridge. Among the polite chatter, one Muriel Bristol—a psychologist studying the mechanisms by which algae acquire nutrients—mentions she has a preference for tea poured over milk, as opposed to milk poured over tea. In a classic example of women not being able to express even the most insignificant preference without an opinionated man telling them they’re wrong, Ronald A. Fisher, a local statistician (later turned eugenicist who dismissed the notion of smoking cigarettes being dangerous as ‘propaganda’, mind you) decides to put her claim to the test with an experiment. Bristol is given eight cups of tea and asked to classify them as milk first or tea first. Luckily, she correctly identifies all eight of them, and gets to happily continue about her life (presumably until the next time she dares mention a similarly outrageous and consequential opinion like a preferred toothpaste brand or a favourite method for filing papers). Fisher, on the other hand, is incentivized to develop Fisher’s exact test, a statistical significance test used in the analysis of contingency tables.
Continue readingPitfalls of using Pearson’s correlation for comparing model performance
Pearson’s R (correlation coefficient) is a measure of the linear correlation between two variables, giving a value between -1 and 1, where 1 is total positive linear correlation, 0 is no linear correlation, and -1 is total negative linear correlation. While it’s a useful statistic for understanding the relationship between two variables, it is often used to compare the performance of two or more models. For example, imagine we had experimental values that we are predicting and several models’ predictions. Obviously, we would prefer the model with the highest Pearson’s R … or perhaps not?
Continue readingOptimising for PR AUC vs ROC AUC – an intuitive understanding
When training a machine learning (ML) model, our main aim is usually to get the ‘best’ model out the other end in an unbiased manner. Of course, there are other considerations such as quick training and inference, but mostly we want to be good at predicting the right answer.
A number of factors will affect the quality of our final model, including the chosen architecture, optimiser, and – importantly – the metric we are optimising for. So, how should we pick this metric?
Continue readingTaking Equivariance in deep learning for a spin?
I recently went to Sheh Zaidi‘s brilliant introduction to Equivariance and Spherical Harmonics and I thought it would be useful to cement my understanding of it with a practical example. In this blog post I’m going to start with serotonin in two coordinate frames, and build a small equivariant neural network that featurises it.
Continue readingThe Surprising Shape of Normal Distributions in High Dimensions
Multivariate Normal distributions are an essential component of virtually any modern deep learning method—be it to initialise the weights and biases of a neural network, perform variational inference in a probabilistic model, or provide a tractable noise distribution for generative modelling.
What most of us (including—until very recently—me) aren’t aware of, however, is that these Normal distributions begin to look less and less like the characteristic bell curve that we associate them with as their dimensionality increases.
Continue readingA Simple Way to Quantify the Similarity Between Two Sets of Molecules
When designing machine learning algorithms with the aim of accelerating the discovery of novel and more effective therapeutics, we often care deeply about their ability to generalise to new regions of chemical space and accurately predict the properties of molecules that are structurally or functionally dissimilar to the ones we have already explored. To evaluate the performance of algorithms in such an out-of-distribution setting, it is essential that we are able to quantify the data shift that is induced by the train-test splits that we rely on to decide which model to deploy in production.
For our recent ICML 2023 paper Drug Discovery under Covariate Shift with Domain-Informed Prior Distributions over Functions, we chose to quantify the distributional similarity between two sets of molecules through the Maximum Mean Discrepancy (MMD).
Continue readingggPlotting tips with OPIG data
Ever wondered whether opiglet keep their ketchup in the fridge or cupboard. Perhaps you’ve wanted to know how to create nice figure to display lots of information simultaniously. Publication quality figures are easy within R with the ggplot package. We may also learn some good visualisation.
Continue reading