Yashar Niknafs, PhD, is on a crusade to explore cancer gene expression datasets-information essential to identify biomarkers that can aid physicians in selecting optimal treatments for patients. He is painstakingly curating and analyzing data from over 20,000 cancer samples into a recently developed, publicly available cancer gene expression data portal called MiPanda (Michigan Portal for the Analysis of NGS Data).
Now settled into a career path with positive implications for the entirety of oncology, Niknafs' earliest years in medicine were a delivery on expectations. It all began in childhood when, every morning before school, a young Niknafs was confronted by a math problem scrawled across a white board by his father, an Iranian immigrant.
"I was supposed to have the problem solved by evening. Dad was very deliberate about teaching me math," recalled Niknafs. "As a result, I always felt like a disciple of science and mathematics. I grew up feeling indebted to the sacrifice my parents had made by leaving their home in Tehran, their family, and their country. It is a stereotype, but one that is not untrue: immigrants from that area of the world expect their children to be successful. My dad wanted me to be a doctor and he expected me to excel. So I did it; I got into med school, but I always felt there was more I could do."
Niknafs entered an MD/PhD program at the University of Michigan, earned his PhD in May 2017, and is currently in a period of postdoctoral research. He describes this period fondly: "I found myself on a fast-moving train; there were few chances to slow down and really ask myself what I want my career to look like. I spent last year growing and finding myself."
Self-Discovery
Asked what he found during his self-discovery year, he answered, "A lot. A big piece was learning compassion for myself, embracing who I am, and allowing my career and my life to be flexible and different from what I always expected them to be."
Niknafs decided to stay in Michigan to do postdoctoral work before returning to medical school to complete his MD. "But a lot of doors opened around me that I had never considered entering before because I was always walking toward a defined endpoint of being a physician. There has been a huge reversal; I do not think I will do a residency because I have been pulled toward science instead. That was a big revelation, and it led me to finding a deeper passion for my work by asserting myself in my life."
The new paths of opportunity took root in 2011 when Niknafs did a rotation in the lab of Arul M. Chinnaiyan, MD, PhD, a Professor of Pathology and Urology at University of Michigan Medical School in Ann Arbor. His lab maintains a large bioinformatic research component, which was a far cry from the wet labs in which Niknafs had previously worked. It proved to be an exciting new frontier.
"When I started, I didn't know how to write a single line of code, but I taught myself how and went from there. Now I code full-time as a bioinformatician. It was a big switch," said Niknafs.
"Most of my work is focused on processing next-gen sequencing data and transcriptomics. Some of my early work involved investigating the space of the non-coding genome-the parts that do not make protein."
He noted that one revelation of the Human Genome Project was that only about 2 percent of the genome makes protein. Once that was discovered, there was an effort afoot to understand what the other 98 percent was actually doing. About 15 years ago, people began to identify non-coding functional RNA; microRNAs (mRNAs) have been studied for years.
"With the advent of next-gen sequencing technology, scientists started to profile the genome on a large scale and realized there were markers for transcription that were widespread in areas where proteins weren't made," Niknafs explained. "It was the advent of RNA sequencing that enabled comprehensive RNA profiling in a cell, which gave us the definitive fact that there are so-called non-coding RNAs all throughout the genome.
"Long non-coding RNAs (lncRNAs) are RNAs that look like mRNAs-they are capped, spliced, polyadenylated, and multi-exonic, but they just don't have the open reading frame needed to make a protein. The focus of much of my work was on discovering lncRNAs and figuring out what they do," he continued. "Why do they exist? A lot of skeptics think they are just 'noise'-leaky transcription. But now there are many anecdotal studies reporting functionality of lncRNAs. Some of them must have purpose. We wanted to find them all and try to facilitate more widespread characterization."
Studies Carry Impact
Niknafs said when he started in the bioinformatics lab "...the goal was to apply bioinformatics to all of the available RNA-seq data we could get our hands on and then use some tools we had built bioinformatically to leverage the data to discover novel non-coding RNAs in various cancer types. We called this the MiTranscriptome. That was a landmark lncRNA study for us (Nat Genet 2015;47(3):199-208)."
To delineate genome-wide lncRNA expression, Niknafs and colleagues curated 7,256 RNA sequencing libraries from tumors, normal tissues, and cell lines.
"The work paid dividends," said Niknafs. "We found a large part of the human genome has the capacity to be transcribed into RNA. Before our work, there had been some efforts to identify lncRNAs using a handful of RNA-sequencing libraries and a small number of tissues. At that time, people thought there were about 10,000 non-coding RNAs compared to the 20,000-30,000 protein-coding genes that are known to be in the genome. But in our study, we identified that there are upward of 50,000 non-coding RNAs spread all throughout the human genome. That was an important discovery." So important, in fact, that the paper has already been cited over 500 times by other authors and studies.
From there, Niknafs facilitated a number of studies investigating the function of some of the more "exciting" lncRNAs discovered in the transcriptome, including studies of lncRNAs in prostate and breast cancers.
Another long-term study looked at highly conserved lncRNAs, including one which Niknafs named THOR (which stands for Testis-associated Highly-conserved Oncogenic long non-coding RNA), found in the genomes of humans, mice, and zebrafish (Cell 2017;171(7):1559-1572.e20). It is unusual for this type of RNA to be conserved throughout several species, leading investigators to believe that if the RNA plays a role in other animals and species besides humans it must be important. THOR is highly expressed in testes cells, yet has little to no expression in other types of adult normal tissue.
In addition to finding THOR expression in normal testis tissue, the researchers found it was highly expressed in some subsets of cancers, particularly lung cancer and melanoma. THOR's expression has a direct impact on cancer development. According to information provided by the University of Michigan, Niknafs' team found that, if they knocked down THOR in cell lines expressing it, tumor growth slowed. If they overexpressed THOR, cells grew faster. And when they eliminated THOR from normal cells, the cells continued to develop normally, suggesting it only impacts cancer cells. THOR could be a good target for drug development because blocking it does not impact normal cells.
"To those who say lncRNAs do nothing, this provides a strong rebuttal," proclaimed Niknafs.
Another effort of which Niknafs is particularly proud is the development of a tool he calls TACO (Transcriptome Assemblies Combined into One), a bioinformatics tool that facilitates transcript structure prediction from large RNA-seq datasets. The standard tools before TACO suffered from poor accuracy in predicting transcript structure. The tool Niknafs and his team built set a new standard for transcript structure prediction from large RNA-seq datasets.
Moving Forward
Asked to characterize his work as he moves forward, Niknafs said, "My premiere work now-and what I hope to use as a springboard to the rest of my career-is the collection, analysis, and processing of data to form an extremely comprehensive transcriptome database. As I was nearing the end of my PhD, I realized I had all this data and I kept getting requests from people asking, 'Can you tell me if this gene is in this cancer?' or 'Can you tell me if these genes correlate to this other gene in breast cancer?' Requests just kept coming in. So I realized if I could build a tool to allow access to this information it would help everyone."
It was a huge endeavor, in light of the fact that Niknafs first had to teach himself web programming and how to build a website. ("But I like learning new things, so it was fun.") He now has a beta version available of the MiPanda tool. While a skeleton of the beta version is available now at http://www.mipanda.org, Niknafs hopes to launch a more useable first generation of the site in next few months.
"Even now, people are using it to search for a gene, and see the expression in 10,000 normal samples, 10,000 tumor samples, and a thousand cell lines. And correlate genes to other genes. Now I am building in clinical outcomes as well," he detailed. "My goal is to build a resource that will help a lot of people do cancer research, and research beyond cancer as well. I am hoping to make a tool that will be useful for all scientists. I call it bridging the gap between data and its usability.
"There is a lot of data available, but in order to use it you need an extensive bioinformatics toolkit, a lot of computational power, and immense data storage; I have the skillset, and Michigan has the computing power and the storage. We are pulling all those things together to enable people to take advantage of the data."
Niknafs, a bachelor who hopes to settle down and have a family one day, diffuses everyday pressures by maintaining two large aquariums ("...there is a certain peacefulness that I enjoy..."), and finding self-expression in a rather unusual undertaking: cocktail mixology.
He explained, "It's just fun. I make my own bitters which require a bittering agent-always a root or bark to give it that bitter characteristic. Then they need a flavoring agent-spices like cinnamon or cardamom-then often an herbal or floral component for the aroma. The components are steeped in high-proof alcohol for a month or two before they are ready for use. Then I mix various ones together in different rations for various flavors. For my signature bitters-cherry vanilla-I bought Michigan cherries, muddled them up, steeped them for a month, then added pecans, cinnamon, cardamom, anise, and vanilla bean. Delicious." And why not? The ingredient amounts and pairings, and the melding of flavors, are simply another arm of mathematics and science.
Valerie Neff Newitt is a contributing writer.
Spotlight on Young Investigators