# So much to do, so little time

Trying to squeeze sense out of chemical data

## Ridit Analysis

While preparing material for an R workshop I ran at NCATS I was pointed to the topic of ridit analysis. Since I hadn’t heard of this technique before I decided to look into it and investigate how R could be used for such an analysis (and yes, there is a package for it).

## Why ridit analysis?

First, lets consider why one might consider ridit analysis. In many scenarios one might have data that is categorical, but the categories are ordered. This type of data is termed ordinal (sometimes also called “nominal with order”). An example might be a trial of an analgesics ability to reduce pain, whose outcome could be no pain, some pain, extreme pain. While there are three categories, it’s clear that there is an ordering to them. Analysis of such data usually makes use of methods devised for categorical data – but such methods will not make use of the information contained within the ordering of the categories. Alternatively, one might numerically code the groups using 1, 2 and 3 and then apply methods devised for continuous or discrete variables. This is not appropriate since one can change the results by simply changing the category coding.

Ridit analysis essentially transforms ordinal data to a probability scale (one could call it a virtual continuous scale). The term actually stands for relative to an identified distribution integral transformation and is analoguous to probit or logit. (Importantly, ridit analysis is closely related to the Wilcoxon rank sum test. As shown by Selvin, the Wilcoxon test statistic and the mean ridit are directly related).

## Definitions

Essentially, one must have at least two groups, one of which is selected as the reference group. Then for the non-reference group, the mean ridit is an

estimate of the probability that a random individual from that group will have a value on the underlying (virtual) continuous scale greater than or equal to the value for a random individual from the reference group.

So if larger values of the underlying scale imply a worse condition, then the mean ridit is the probability estimate that the random individual from the group is worse of than a random individual from the reference group (based on the interpretation from Bross). Based on the definition of a ridit (see here or here), one can compute confidence intervals (CI) or test the hypothesis that different groups have equal mean ridits. Lets see how we can do that using R

## Mechanics of ridit analysis

Consider a dataset taken from Donaldson, (Eur. J. Pain, 1988) which looked at the effect of high and low levels of radiation treatment on trials participants’s sleep. The numbers are counts of patients:

 123456789101112 sleep <- data.frame(pain.level=factor(c('Slept all night with no pain',                       'Slept all night with some pain',                       'Woke with pain - medication provided relief',                       'Woke with pain - medication provided no relief',                       'Awake most or all of night with pain'),                       levels=c('Slept all night with no pain',                         'Slept all night with some pain',                         'Woke with pain - medication provided relief',                         'Woke with pain - medication provided no relief',                         'Awake most or all of night with pain')),                     low.dose=c(3, 10, 6,  2, 1),                     high.dose=c(6,10,2,0,0))

Here the groups are in the columns (low.dose and high.dose) and the categories are ordered such tat Awake most or all of night with pain is the “maximum” category. To compute the mean ridits for each dose group we first reorder the table and then convert the counts to proportions and then compute ridits for each category (i.e., row).

 1234567891011121314151617 ## reorder table sleep <- sleep[ length(levels(sleep$pain.level)):1, ] ## compute proportions sleep$low.dose.prop <- sleep$low.dose / sum(sleep$low.dose) sleep$high.dose.prop <- sleep$high.dose / sum(sleep$high.dose) ## compute riddit ridit <- function(props) { ## props should be in order of levels (highest to lowest) r <- rep(-1, length(props)) for (i in 1:length(props)) { if (i == length(props)) vals <- 0 else vals <- props[(i+1):length(props)] r[i] <- sum(vals) + 0.5*props[i] } return(r) } sleep$low.dose.ridit <- ridit(sleep$low.dose.prop) sleep$high.dose.ridit <- ridit(sleep$high.dose.prop) The resultant table is below  123456 pain.level low.dose high.dose low.dose.prop high.dose.prop low.dose.ridit high.dose.ridit 5 Awake most or all of night with pain 1 0 0.04545455 0.0000000 0.97727273 1.0000000 4 Woke with pain - medication provided no relief 2 0 0.09090909 0.0000000 0.90909091 1.0000000 3 Woke with pain - medication provided relief 6 2 0.27272727 0.1111111 0.72727273 0.9444444 2 Slept all night with some pain 10 10 0.45454545 0.5555556 0.36363636 0.6111111 1 Slept all night with no pain 3 6 0.13636364 0.3333333 0.06818182 0.1666667 The last two columns represent the ridit values for each category and can be interpreted as a probability estimate that an individuals value on the underlying continuous scale is less than or equal to the midpoint of the corresponding interval The next step (and main point of the analysis) is to compute the mean ridit for a group (essentially the sum of the category proportions for that group weighted by the category ridits in the reference group) , based on a reference. In this case, lets assume the low dose group is the reference.  1 mean.r.high <- sum(with(sleep, high.dose.prop * low.dose.ridit)) which is 0.305, and can be interpreted as the probability that a patient receiving the high dose of radiation will experience more sleep interference than a patient in the low dose group. Importantly, since ridits are estimates of probabilities, the complementary ridit (i.e., using the high dose group as reference) comes out to 0.694 and is the probability that a patient in the low radiation dose group will experience more sleep interferance than a patient in the high dose group. ## Statistics on ridits There are a number of ways to compute CI’s on mean ridits or else test the hypothesis that the mean ridits differ between $k$ groups. Donaldsons method for CI calculation appears to be restricted to two groups. In contrast, Fleiss et al suggest an alternate method based on. Considering the latter, the CI for a group vs the reference group is given by $\overline{r}_i \pm B \frac{\sqrt{n_s +n_i}}{2\sqrt{3 n_s n_i}}$ where $\overline{r}_i$ is the mean ridit for the $i$‘th group, $n_s$ and $n_i$ are the sizes of the reference and query groups, respectively and $B$ is the multiple testing corrected standard error. If one uses the Bonferroni correction, it would be $1.96 \times 1$ since there is only two groups being compared (and so 1 comparison). Thus the CI for the mean ridit for the low dose group, using the high dose as reference is given by $0.694 \pm 1.96 \frac{\sqrt{18 + 22}}{2\sqrt{3 \times 18 \times 22}}$ which is 0.515 to 0.873. Given that the interval does not include 0.5, we can conclude that there is a statistically significant difference ($\alpha = 0.05$) in the mean ridits between the two groups. For the case of multiple groups, the CI for any group vs any other group (i.e., not considering the reference group) is given by $(\overline{r}_i - \overline{r}_j + 0.5) \pm B \frac{\sqrt{n_i + n_j}}{2\sqrt{3 n_i n_j}}$ Fleiss et al also describes how one can test the hypothesis that the mean ridits across all groups (including the reference) are equal using a $\chi^2$ statistic. In addition, they also describe how one can perform the same test between any group and the reference group. ## R Implementation I’ve implemented a function that computes mean ridits and their 95% confidence interval (which can be changed). It expects that the data is provided as counts for each category and that the input data.frame is ordered in descending order of the categories. You need to specify the variable representing the categories and the reference variable. As an example of its usage, we use the dataset from Fleiss et al which measured the degree of pain relief provided by different drugs after oral surgery. We perfrom a ridit analysis using aspirin as the reference group:  12345678 dental <- data.frame(pain.relief = factor(c('Very good', 'Good', 'Fair', 'Poor', 'None'), levels=c('Very good', 'Good', 'Fair', 'Poor', 'None')), ibuprofen.low = c(61, 17, 10, 6, 0), ibuprofen.high = c(52, 25, 5, 3, 1), Placebo = c(32, 37, 10, 18, 0), Aspirin = c(47, 25, 11, 4, 1) ) ridit(dental, 'pain.relief', 'Aspirin') which gives us  1234567891011121314151617$category.ridit   pain.relief ibuprofen.low ibuprofen.high    Placebo 1   Very good    0.67553191    0.697674419 0.83505155 2        Good    0.26063830    0.250000000 0.47938144 3        Fair    0.11702128    0.075581395 0.23711340 4        Poor    0.03191489    0.029069767 0.09278351 5        None    0.00000000    0.005813953 0.00000000 $mean.ridit ibuprofen.low ibuprofen.high Placebo 0.5490812 0.5455206 0.3839620$ci            group       low      high 1  ibuprofen.low 0.4361128 0.6620496 2 ibuprofen.high 0.4300396 0.6610016 3        Placebo 0.2718415 0.4960826

The results suggest that patients receiving either dose of ibuprofen will get better pain relief compared to aspirin. However, if you consider the CI’s it’s clear that they both contain 0.5 and thus there is no statistical difference in the mean ridits for these two doses, compared to aspirin. On the other hand, placebo definitely leads to less pain relief compared to aspirin.

Written by Rajarshi Guha

January 18th, 2015 at 7:26 pm

## Summarizing Collections of Curves

I was browsing live notes from the recent IEEE conference on visualization and came across a paper about functional boxplots. The idea is an extension of the boxplot visualization (shown alongside), to a set of functions. Intuitively, one can think of a functional box plot as specific envelopes for a set of functions. The construction of this plot is based on the notion of band depth (see the more general concept of data depth) which is a measure of how far a given function is from the collection of functions. As described in Sun & Genton the band depth for a given function can be computed by randomly selecting $J$ functions and identifying wether the given function is contained within the minimum and maximum of the $J$ functions. Repeating this multiple times, the fraction of times that the given function is fully contained within the $J$ random functions gives the band depth, $BD_j$. This is then used to order the functions, allowing one to compute a 50% band, analogous to the IQR in a traditional boxplot. There are more details (choice of $J$, partial bounding, etc.) described in the papers and links above.

My interest in this approach was piqued since one way of summarizing a dose response screen, or comparing dose response data across multiple conditions is to generate a box plot of a single curve fit parameter – say, $\log IC_{50}$. But if we wanted to consider the curves themselves, we have a few options. We could simply plot all of them, using translucency to avoid a blob. But this doesn’t scale visually. Another option, on the left, is to draw a series of box plots, one for each dose, and then optionally join the median of each boxplot giving a “median curve”. While these vary in their degree of utility, the idea of summarizing the distribution of a set of curves, and being able to compare these distributions is attractive. Functional box plots look like a way to do this. (A cool thing about functional boxplots is that they can be extended to multivariate functions such as surfaces and so on. See Mirzargar et al for examples)

Computing $BD_j$ can be time consuming if the number of curves is large or $J$ is large. Lopez-Pintado & Jornsten suggest a simple optimization to speed up this step, and for the special case of $J = 2$, Sun et al proposed a ranking based procedure that scales to thousands of curves. The latter is implemented in the fda package for R which also generates the final functional box plots.

As an example I considered 6 cell proliferation assays run in dose response, each one running the same set of compounds, but under different growth conditions. For each assay I only considered good quality curves (giving from 349 to 602 curves). The first plot compares the actives identified in the different growth conditions using the $\log IC_{50}$, and indicates a statistically significant increase in potency in the last three conditions compared to the first three.

In contrast, the functional box plots for the 6 assays, suggest a somewhat different picture (% Response = 100 corresponds to no cell kill and 0 corresponds to full cell kill).

The red dashed curves correspond to outliers and the blue lines correspond to the ‘maximum’ and ‘minimum’ curves (analogous to the whiskers of the traditional boxplot). Importantly, these are not measured curves, but instead correspond to the dose-wise maximum (and minimum) of the real curves. The pink region represents 50% of the curves and the black line represents the (virtual) median curve. In each case the X-axis corresponds to dose (unlabeled to save space). Personally, I think this visualization is a little cleaner than the dose-wise box plot shown above.

The mess of red lines in the plot 1 suggest an issue with the assay itself. While the other plots do show differences, it’s not clear what one can conclude from this. For example, in the plot for 4, the dip on the left hand side (i.e., low dose) could suggest that there is a degree of cytotoxicity, which is comparatively less in 3, 5 and 6. Interestingly none of the median curves are really sigmoidal, suggesting that the distribution of dose responses has substantial variance.

Written by Rajarshi Guha

November 30th, 2014 at 3:02 pm

## Thoughts on the DREAM Synergy Prediction Challenge

The DREAM consortium has run a number of predictive modeling challenges and the latest one on predicting small molecule synergies has just been published. The dataset that was provided included baseline gene expression of the cell line (OCI-LY3), expression in presence of compound (2 concentrations, 2 time points), dose response data for 14 compounds and the excess over Bliss for the 91 pairs formed from the 14 compounds. Based on this data (and available literature data) participants had to predict a ranking for the 91 combinations.

The paper reports the results of 31 approaches (plus one method that was not compared to the others) and does a good job of summarizing their performance and identifying whether certain data type or certain approaches work better than others. They also investigated the performance of an ensemble of approaches, which, as one might expect, worked better than the single methods. While the importance of gene expression in predictive performance was not as great as I would’ve thought, it was certainly more useful than chemical structure alone. Interestingly, they also noted that “compounds with more targeted mechanisms, such as rapamycin and blebbistatin, were least synergistic“. I suspect that this is somewhat dataset specific, but it will be interesting to see whether this holds in large collections of combination experiment such as those run at NCATS.

Overall, it’s an important contribution with the key take home message being

… synergy and antagonism are highly context specific and are thus not universal properties of the compounds’ chemical, structural or substrate information. As a result, predictive methods that account for the genetics and regulatory architecture of the context will become increasingly relevant to generalize results across multiple contexts

Given the relative dearth of predictive models of compound synergy, this paper is a nice compilation of methods. But there are some issues that weaken the paper.

• One key issue are the conclusions on model performance. The organizers defined a score, termed probabilistic c-score (PC score). If I understand correctly, a random ranking should give PC = 0.5. It turns out that the best performing method exhibited a PC score = 0.61 with a number of methods hovering around 0.5. Undoubtably, this is a tough problem, but when the authors states that “… this challenge shows that current methodologies can perform significantly better than chance …” I raise an eyebrow. I can only assume that what they meant was that the results were “statistically significantly better than chance“, because in terms of effect size the results are not impressive. After reading this excellent article on p-values and significance testing I’m particularly sensitized to claims of significance.
• The dataset could have been strengthened by the inclusion of self-crosses. This would’ve allowed the authors to assess actual excess over Bliss values corresponding to additivity (which will not be exactly 0 due to experimental noise), and avoid the use of cutoffs in determining what is synergistic or antagonistic.
• Similarly, a key piece of data that would really strengthen these approaches is the expression data in presence of combinations. While it’s unreasonable to have this data available for all combinations, it could be used as a first step in developing models to predict the expression profile in presence of combination treatment. Certainly, such data could be used to validate some assumptions made by some of the models described (e.g., concordance of DEG’s induced by single agents implies synergistic response).
• Kudos for including source code for the top methods, but would’ve been nicer if data files were included so we could actually reproduce the results.
• The authors conclude that when designing new synergy experiments, one should identify mechanistically diverse molecules to make up for the “small number of potentially synergistic pathways“. While mechanistic diversity is a good idea, it’s not clear how they conclude there are a small number of pathways that play a role in synergy.
• It’s a pity that the SynGen method was not compared to the other methods. While the authors provide a justification, it seems rather weak. The method only applied to the synergistic combinations (performance was not a whole lot better than random – true positive rate of 56%) – but the text indicates that it predicted synergistic compound pairs. It’s not clear whether this means it made a call on synergy or a predicted ranking. If the latter it would’ve been interesting to see how it compared to the rankings of the synergistic subset of 91 compounds from other methods.

Written by Rajarshi Guha

November 20th, 2014 at 5:37 pm

## Metabolite Similarity & Dirty Compounds

with one comment

Edit 10/9/14 – Updated statistics for the 1024 bit fingerprints

There’s been some discussion about a paper by O’Hagan et al that have proposed a Rule of 0.5 that states that 90% of approved drugs exhibit a Tanimoto similarity > 0.5 to one or more human metabolites. Their analysis is based on metabolites listed in Recon2, a reconstruction of the human metabolic network. The idea makes sense and there’s an in depth discussion at In the Pipeline.

Given the authors’ claim that

a successful drug is likely to lie within a Tanimoto distance of 0.5 of a known human metabolite. While this does not mean, of course, that a molecule obeying the rule is likely to become a marketed drug for humans, it does mean that a molecule that fails to obey the rule is statistically most unlikely to do so

I was interested in seeing how this rule of thumb holds up when faced with compounds that are not supposed to make it through the drug development pipeline. Since PAINS appear to be the structural filter du jour, I decided to look at compounds that failed the PAINS filter. I worked with the 10,000 compounds included in Saubern et al. Simon Saubern provided me the set of 861 compounds that failed the PAINS filters, allowing me to extract the set of compounds that passed (9139)

Chris Swain was kind enough to extract the compound entries from the Matlab dump provided by O’Hagan et al. This file contained InChI representations for a subset of the entries. I extracted the 2980 valid InChI strings and converted them to SMILES using ChemAxon molconvert 6.0.5. The processed data (metabolite name, InChI and SMILES) are available here. However, after deduplication, there were 1335 unique metabolites

Now, O’Hagan et al for some reason, used the 166 bit MACCS keys, but hashed them to 1024 bits. Usually, when using a keyed fingerprint, the goal is to retain the correspondence between bit position and substructure. The hashing step results in a loss of such correspondence. So it’s a bit surprising that they didn’t use some sort of path (Daylight) or environment (ECFPn) based fingerprint. Since I didn’t know how they hashed the MACCS keys, I calculated 166 bit MACCS keys and 1024 bt ECFP6 and extended path fingerprints using the CDK (via rcdk). Then for each compound in the PAINS pass or fail set, I computed the similarity to each of the 1335 metabolites and identified the maximum similarity (termed NMTS in the paper) and then plotted the distribution of these NMTS values between the PAINS pass and fail sets.

First, the similarity cutoff proposed by the authors is obiously dependent on the fingerprint. So while the bulk of the 166 bit MACCS similarities are > 0.5, this is not really meaningful. A more relevant comparison is to 1024 bit fingerprints – both are hashed, so should be somewhat comparable to the authors choice of hashed MACCS keys.

The path fingerprints lead to an NMTS of ~ 0.25 for both PAINS pass and fail sets and the ECFP6 leads to an NMTS of ~ 0.18 for both sets. Though the difference in medians between the pass and fail sets for the path fingerprint is statistically significant (p = 1.498e-05, Wilcoxon test), the difference itself is very small: 0.005. (For the circular fingerprint there is no statistically significant difference). However, the PAINS pass set does contain more outliers with values > 0.5. In that sense the proposed rule does separate the two groups. Of the top of my head I don’t know whether the WEHI screening deck that was the source of the 10,000 compounds was designed to be drug-like. At the same time all this might be saying is there is no relationship between metabolite-likenes and PAINS-likeness.

It’d be interesting to see how this type of analysis holds up with other well known filter rules (REOS, Lilly etc). A related thing to look at would be to see how druglikeness scores compare with NMTS values.

Code and data are available in this repository

Written by Rajarshi Guha

October 7th, 2014 at 5:47 pm

## Predicting Synergy from Lipophilicity?

I recently came across a paper by Yilancioglu et al that described a method to predict drug synergies using only lipophilicity. In effect, it claimed to predict synergy based purely on a physicochemical property and independent of target or pathway information. Their results suggest that

combinations of two lipophilic drugs had a greater tendency to show drug synergy

I must admit that I’m skeptical of this claim. While lipophilicity certainly plays a role in a drugs effect (and thereby a drug combinations’ effect), I’m not sure that lipophilicty is a primary driver of a synergistic interaction. Rather, lipophilicity might be a prerequisite; that is, if two molecules cannot enter the cell to access their target(s), they’re unlikely to exhibit synergy!

The paper considered a set of 175 (anti-fungal) drug pairs tested in yeastand evaluated molecular weight, logP, H-bond donor and acceptor counts and also computed a synergicity, that is a measure of how frequently a drug exhibits synergy with other drugs. So the work isn’t really directly capturing synergy (which was measured using the Loewe model). They they compute Spearman correlation between the synergicity and the various physicochemical properties – identifying logP as the one with a statistically significant, though moderate correlation (though one of their examples presents a significant correlation of 0.2 – not a whole lot you could do with that!). They then go on to build a decision tree model that predicts synergicity surprisingly well, though given that the model is based on a synergy netowrk (nodes are drugs, edges are weighted by the synergy between a pair of drugs), it’s not clear how they evaluated the lipophilicity of a drug pair. The terminology was a bit confusing – sometimes using synergicity and sometimes synergy. It’s definitely a surprising result – but is it really meaningful? As I note above, I find it difficult to accept lipophilicty as a proximal driver of synergy. The fact that one of their analyses employs binned logP could raise an issue (see a presentation or this paper on the dangers of binned data).

Given that NCATS has developed a high throughput compound combination screening platform, I was interested in seeing if any of this held up on some of our public datasets. I considered a dataset of 466 drugs tested in combination (6×6 matrix) with Ibrutinib. Thus, in contrast to the Yilancioglu et al paper, one member of the combinations is constant. As a result, it makes sense to correlate the logP of the other (i.e., non Ibrutinib) component of the combination to the synergy value of that combination. I evaluated logP of the compounds using ChemAxons cxcalc tool and compared the values to the various synergy metrics we calculate (see here for definitions).

The figure above pretty much shows no correlation for any of the synergy metrics (and Spearmans ρ was ≅ 0 with p > 0.05). I also repeated the calculation of a set of 1912 combinations (i.e., 1912 compounds combined with Ibrutinib) and got essentially the same result. Granted that this was on a single lymphoma cell line (TMD8) which is significantly different from the environment considered by the authors and that our synergy metrics are different from those described in the paper. So, it might just be a feature of anti-fungal drugs?

But interestingly, when we considered binned logP values and look at the median value of a synergy metric in each logP bin, we do see a trend – at least for two out of four metrics. But given the scatter plots, where the variability is not hidden, is this really meaningful?

So overall, the paper presents some surprising observations but is a little unsatisfying from an explanatory point of view. And the conclusions don’t seem to translate to other datasets.

Written by Rajarshi Guha

July 31st, 2014 at 4:01 am

Posted in research